R
@@ -1116,7 +1116,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -1349,7 +1349,7 @@
Challenge
-
+
R
@@ -1625,7 +1625,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1920,7 +1920,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -2090,7 +2090,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
@@ -2375,7 +2375,7 @@
Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -2528,7 +2528,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -2821,7 +2821,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -3010,7 +3010,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -3019,231 +3021,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -3287,7 +3081,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -3436,7 +3230,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -3453,7 +3247,7 @@
Give me a hint
-
+
TODO
@@ -3482,7 +3276,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -3496,7 +3290,7 @@
Give me a hint
-
+
TODO
@@ -3526,7 +3320,7 @@ Exercise 3: Workflow
-
+
TODO
@@ -3598,7 +3392,7 @@ Key Points
-->
Content from Multi-sample analyses
-Last updated on 2024-09-06 |
+
Last updated on 2024-09-08 |
Edit this page
@@ -3655,13 +3449,23 @@ Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.
We start by loading the data and doing a quick exploratory analysis,
essentially applying the normalization and visualization techniques that
-we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this
+time we’re selecting samples 5 to 10, not just 5 by itself.
R
library(MouseGastrulationData)
-sce <- WTChimeraData(samples=5:10, type = "processed")
+library(batchelor)
+library(edgeR)
+library(scater)
+library(ggplot2)
+library(scran)
+library(pheatmap)
+library(scuttle)
+
+sce <- WTChimeraData(samples = 5:10, type = "processed")
+
sce
@@ -3736,15 +3540,12 @@ OUTPUT<
R
-library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3752,23 +3553,24 @@ R
sce <- sce[,idx]
-We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps,
+and visualize them in a tSNE plot.
R
-# normalization
-sce <- logNormCounts(sce)
+sce <- logNormCounts(sce)
+
+dec <- modelGeneVar(sce, block = sce$sample)
-# identify highly variable genes
-dec <- modelGeneVar(sce, block=sce$sample)
chosen.hvgs <- dec$bio > 0
-# dimensionality reduction
sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000)
+
sce <- runTSNE(sce, dimred = "PCA")
sce$sample <- as.factor(sce$sample)
+
plotTSNE(sce, colour_by = "sample")
Challenge
-
+
Samples 5 and 6 were from the same “pool” of cells. Looking at the
documentation for the dataset under ?WTChimeraData we see
@@ -3816,7 +3618,7 @@
Show me the solution
first TSNE plot. If you were developing single-cell library preparation
protocols you might want to preserve this effect to understand how
variation in pools leads to variation in expression, but for now, given
-that we’re investigate other effects, we’ll want to remove this as
+that we’re investigating other effects, we’ll want to remove this as
undesired technical variation.
@@ -3825,29 +3627,30 @@ Show me the solution
Correcting batch effects
-We correct the effect of samples by aid of the
+
We “correct” the effect of samples with the
correctExperiment function in the batchelor
-package and using the sample colData column as
-batch.
+package and using the sample column as batch.
R
-library(batchelor)
-set.seed(10102)
-merged <- correctExperiments(sce,
- batch=sce$sample,
- subset.row=chosen.hvgs,
- PARAM=FastMnnParam(
- merge.order=list(
+set.seed(10102)
+
+merged <- correctExperiments(
+ sce,
+ batch = sce$sample,
+ subset.row = chosen.hvgs,
+ PARAM = FastMnnParam(
+ merge.order = list(
list(1,3,5), # WT (3 replicates)
list(2,4,6) # td-Tomato (3 replicates)
)
)
)
-merged <- runTSNE(merged, dimred="corrected")
-plotTSNE(merged, colour_by="batch")
+merged <- runTSNE(merged, dimred = "corrected")
+
+plotTSNE(merged, colour_by = "batch")
Once we removed the sample batch effect, we can proceed with the
Differential Expression Analysis.
@@ -3869,7 +3672,7 @@ Challenge
-
+
It’s important to have multiple samples within each experimental
group because it helps the batch effect correction algorithm distinguish
@@ -3892,7 +3695,7 @@
Show me the solution
Differential Expression
-In order to perform a Differential Expression Analysis, we need to
+
In order to perform a differential expression analysis, we need to
identify groups of cells across samples/conditions (depending on the
experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell
@@ -3920,9 +3723,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output
# corresponds to one unique combination of these two factors.
-library(scuttle)
-summed <- aggregateAcrossCells(merged,
- id=colData(merged)[,c("celltype.mapped", "sample")])
+
+summed <- aggregateAcrossCells(
+ merged,
+ id = colData(merged)[,c("celltype.mapped", "sample")]
+)
+
summed
R
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
y
@@ -4012,7 +3818,9 @@ R
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)
@@ -4028,8 +3836,10 @@ OUTPUT<
R
-keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)
@@ -4049,6 +3859,7 @@ R
y <- calcNormFactors(y)
+
y$samples
@@ -4089,7 +3900,7 @@ Rpar(mfrow = c(2,3))
for (i in seq_len(ncol(y))) {
- plotMD(y, column=i)
+ plotMD(y, column = i)
}
RR
-limma::plotMDS(cpm(y, log=TRUE),
- col=ifelse(y$samples$tomato, "red", "blue"))
+limma::plotMDS(cpm(y, log = TRUE),
+ col = ifelse(y$samples$tomato, "red", "blue"))
We then construct a design matrix by including both the pool and the
tomato as factors. This design indicates which samples belong to which
@@ -4147,6 +3958,7 @@
R
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)
@@ -4166,7 +3978,7 @@ RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized
-linear model for each gene. The robust=TRUE parameter
+linear model for each gene. The robust = TRUE parameter
avoids distortions from highly variable clusters. The QL method includes
an additional dispersion parameter, useful to handle the uncertainty and
variability of the per-gene variance, which is not well estimated by the
@@ -4176,7 +3988,8 @@
RR
-fit <- glmQLFit(y, design, robust=TRUE)
+fit <- glmQLFit(y, design, robust = TRUE)
+
summary(fit$var.prior)
@@ -4215,7 +4028,8 @@ RR
-res <- glmQLFTest(fit, coef=ncol(design))
+res <- glmQLFTest(fit, coef = ncol(design))
+
summary(decideTests(res))
@@ -4255,14 +4069,14 @@ OUTPUT<
R
-library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)
The returned object is a list of DataFrames each with
@@ -4274,6 +4088,7 @@
R
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]
Challenge
-
+
“logFC” stands for log fold-change. Rather than reporting e.g. a
5-fold increase, it’s better to report a logFC of log(5) = 1.61.
@@ -4330,26 +4145,29 @@
Show me the solution
With DA we look for differences in cluster abundance across
conditions (the tomato injection in our case), rather than differences
in gene expression.
-We first setup some code and variables for further analysis, like
-quantifying the number of cells per each cell type and fit a model to
-catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this
-time we start our analysis on the computed abundances and without
+
Our first steps are quantifying the number of cells per each cell
+type and fitting a model to catch differences between the injected cells
+and the background.
+The process is very similar differential expression modeling, but
+this time we start our analysis on the computed abundances and without
normalizing the data with TMM.
R
-library(edgeR)
-abundances <- table(merged$celltype.mapped, merged$sample)
+abundances <- table(merged$celltype.mapped, merged$sample)
+
abundances <- unclass(abundances)
-# Attaching some column metadata.
+
extra.info <- colData(merged)[match(colnames(abundances), merged$sample),]
-y.ab <- DGEList(abundances, samples=extra.info)
+
+y.ab <- DGEList(abundances, samples = extra.info)
design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples)
-y.ab <- estimateDisp(y.ab, design, trend="none")
-fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)
+
+y.ab <- estimateDisp(y.ab, design, trend = "none")
+
+fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)
Background on compositional effect
@@ -4395,6 +4213,7 @@ R
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
@@ -4407,11 +4226,11 @@ OUTPUT<
R
-y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))
@@ -4427,7 +4246,7 @@ OUTPUT<
R
-topTags(res2, n=10)
+topTags(res2, n = 10)
OUTPUT
@@ -4463,7 +4282,8 @@ Testing against a log-fold
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
@@ -4544,70 +4364,72 @@ OUTPUT<
[8] base
other attached packages:
- [1] edgeR_4.2.0 limma_3.60.2
- [3] batchelor_1.20.0 scran_1.32.0
- [5] scater_1.32.0 ggplot2_3.5.1
- [7] scuttle_1.14.0 MouseGastrulationData_1.18.0
- [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0
-[11] SummarizedExperiment_1.34.0 Biobase_2.64.0
-[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1
-[15] IRanges_2.38.0 S4Vectors_0.42.0
-[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0
-[19] matrixStats_1.3.0 BiocStyle_2.32.0
+ [1] pheatmap_1.0.12 scran_1.32.0
+ [3] scater_1.32.0 ggplot2_3.5.1
+ [5] scuttle_1.14.0 edgeR_4.2.0
+ [7] limma_3.60.2 batchelor_1.20.0
+ [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0
+[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0
+[13] Biobase_2.64.0 GenomicRanges_1.56.0
+[15] GenomeInfoDb_1.40.1 IRanges_2.38.0
+[17] S4Vectors_0.42.0 BiocGenerics_0.50.0
+[19] MatrixGenerics_1.16.0 matrixStats_1.3.0
+[21] BiocStyle_2.32.0
loaded via a namespace (and not attached):
- [1] DBI_1.2.3 formatR_1.14
- [3] gridExtra_2.3 rlang_1.1.3
- [5] magrittr_2.0.3 compiler_4.4.1
- [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0
- [9] png_0.1-8 vctrs_0.6.5
- [11] pkgconfig_2.0.3 crayon_1.5.2
- [13] fastmap_1.2.0 dbplyr_2.5.0
- [15] magick_2.8.3 XVector_0.44.0
- [17] labeling_0.4.3 utf8_1.2.4
- [19] rmarkdown_2.27 UCSC.utils_1.0.0
- [21] ggbeeswarm_0.7.2 purrr_1.0.2
- [23] bit_4.0.5 bluster_1.14.0
- [25] xfun_0.44 zlibbioc_1.50.0
- [27] cachem_1.1.0 beachmat_2.20.0
- [29] jsonlite_1.8.8 blob_1.2.4
- [31] highr_0.11 DelayedArray_0.30.1
- [33] BiocParallel_1.38.0 cluster_2.1.6
- [35] irlba_2.3.5.1 parallel_4.4.1
- [37] R6_2.5.1 Rcpp_1.0.12
- [39] knitr_1.47 splines_4.4.1
- [41] igraph_2.0.3 Matrix_1.7-0
- [43] tidyselect_1.2.1 viridis_0.6.5
- [45] abind_1.4-5 yaml_2.3.8
- [47] codetools_0.2-20 curl_5.2.1
- [49] lattice_0.22-6 tibble_3.2.1
- [51] withr_3.0.0 KEGGREST_1.44.0
- [53] BumpyMatrix_1.12.0 Rtsne_0.17
- [55] evaluate_0.23 BiocFileCache_2.12.0
- [57] ExperimentHub_2.12.0 Biostrings_2.72.1
- [59] pillar_1.9.0 BiocManager_1.30.23
- [61] filelock_1.0.3 renv_1.0.7
- [63] generics_0.1.3 BiocVersion_3.19.1
- [65] sparseMatrixStats_1.16.0 munsell_0.5.1
- [67] scales_1.3.0 glue_1.7.0
- [69] metapod_1.12.0 tools_4.4.1
- [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0
- [73] ScaledMatrix_1.12.0 locfit_1.5-9.9
- [75] cowplot_1.1.3 grid_4.4.1
- [77] AnnotationDbi_1.66.0 colorspace_2.1-0
- [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0
- [81] BiocSingular_1.20.0 vipor_0.4.7
- [83] cli_3.6.2 rsvd_1.0.5
- [85] rappdirs_0.3.3 fansi_1.0.6
- [87] viridisLite_0.4.2 S4Arrays_1.4.1
- [89] dplyr_1.1.4 ResidualMatrix_1.14.0
- [91] gtable_0.3.5 digest_0.6.35
- [93] dqrng_0.4.1 SparseArray_1.4.8
- [95] ggrepel_0.9.5 farver_2.1.2
- [97] rjson_0.2.21 memoise_2.0.1
- [99] htmltools_0.5.8.1 lifecycle_1.0.4
-[101] httr_1.4.7 statmod_1.5.0
-[103] mime_0.12 bit64_4.0.5
+ [1] RColorBrewer_1.1-3 jsonlite_1.8.8
+ [3] magrittr_2.0.3 ggbeeswarm_0.7.2
+ [5] magick_2.8.3 farver_2.1.2
+ [7] rmarkdown_2.27 zlibbioc_1.50.0
+ [9] vctrs_0.6.5 memoise_2.0.1
+ [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1
+ [13] S4Arrays_1.4.1 AnnotationHub_3.12.0
+ [15] curl_5.2.1 BiocNeighbors_1.22.0
+ [17] SparseArray_1.4.8 cachem_1.1.0
+ [19] ResidualMatrix_1.14.0 igraph_2.0.3
+ [21] mime_0.12 lifecycle_1.0.4
+ [23] pkgconfig_2.0.3 rsvd_1.0.5
+ [25] Matrix_1.7-0 R6_2.5.1
+ [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12
+ [29] digest_0.6.35 colorspace_2.1-0
+ [31] AnnotationDbi_1.66.0 dqrng_0.4.1
+ [33] irlba_2.3.5.1 ExperimentHub_2.12.0
+ [35] RSQLite_2.3.7 beachmat_2.20.0
+ [37] filelock_1.0.3 labeling_0.4.3
+ [39] fansi_1.0.6 httr_1.4.7
+ [41] abind_1.4-5 compiler_4.4.1
+ [43] bit64_4.0.5 withr_3.0.0
+ [45] BiocParallel_1.38.0 viridis_0.6.5
+ [47] DBI_1.2.3 highr_0.11
+ [49] rappdirs_0.3.3 DelayedArray_0.30.1
+ [51] rjson_0.2.21 bluster_1.14.0
+ [53] tools_4.4.1 vipor_0.4.7
+ [55] beeswarm_0.4.0 glue_1.7.0
+ [57] grid_4.4.1 Rtsne_0.17
+ [59] cluster_2.1.6 generics_0.1.3
+ [61] gtable_0.3.5 BiocSingular_1.20.0
+ [63] ScaledMatrix_1.12.0 metapod_1.12.0
+ [65] utf8_1.2.4 XVector_0.44.0
+ [67] ggrepel_0.9.5 BiocVersion_3.19.1
+ [69] pillar_1.9.0 BumpyMatrix_1.12.0
+ [71] splines_4.4.1 dplyr_1.1.4
+ [73] BiocFileCache_2.12.0 lattice_0.22-6
+ [75] renv_1.0.7 bit_4.0.5
+ [77] tidyselect_1.2.1 locfit_1.5-9.9
+ [79] Biostrings_2.72.1 knitr_1.47
+ [81] gridExtra_2.3 xfun_0.44
+ [83] statmod_1.5.0 UCSC.utils_1.0.0
+ [85] yaml_2.3.8 evaluate_0.23
+ [87] codetools_0.2-20 tibble_3.2.1
+ [89] BiocManager_1.30.23 cli_3.6.2
+ [91] munsell_0.5.1 Rcpp_1.0.12
+ [93] dbplyr_2.5.0 png_0.1-8
+ [95] parallel_4.4.1 blob_1.2.4
+ [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2
+ [99] scales_1.3.0 purrr_1.0.2
+[101] crayon_1.5.2 rlang_1.1.3
+[103] cowplot_1.1.3 KEGGREST_1.44.0
+[105] formatR_1.14
Exercises
@@ -4630,7 +4452,7 @@ Exercise 1: Replicates
-
+
Remember, you can subset SingleCellExperiments with logical indices,
just like a matrix. You can also access their column data with the
@@ -4644,7 +4466,7 @@
Give me a hint
-
+
R
@@ -4652,11 +4474,12 @@ R
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
@@ -4680,7 +4503,7 @@ Exercise 2:
-
+
You can just hand pheatmap() a matrix as its only
argument. It has a million options, but the defaults are usually pretty
@@ -4694,15 +4517,13 @@
Give me a hint
-
+
R
-library(pheatmap)
-
-pheatmap(y.ab$counts)
+pheatmap(y.ab$counts)
The top DA result was a decrease in ExE ectoderm in the tomato
condition, which you can sort of see, especially if you
@@ -6462,7 +6283,7 @@
Exercise 1
-
+
R
@@ -6495,7 +6316,7 @@ Exercise 2
-
+
R
@@ -6527,7 +6348,7 @@ Exercise 3
-
+
R
@@ -6568,7 +6389,7 @@ Exercise 4
-
+
R
diff --git a/cell_type_annotation.html b/cell_type_annotation.html
index 067a8e8..a423c5a 100644
--- a/cell_type_annotation.html
+++ b/cell_type_annotation.html
@@ -515,7 +515,7 @@ Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -666,7 +666,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -956,7 +956,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -1144,7 +1144,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -1153,231 +1155,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -1421,7 +1215,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -1566,7 +1360,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -1583,7 +1377,7 @@
Give me a hint
-
+
TODO
@@ -1612,7 +1406,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -1626,7 +1420,7 @@
Give me a hint
-
+
TODO
@@ -1656,7 +1450,7 @@ Exercise 3: Workflow
-
+
TODO
diff --git a/eda_qc.html b/eda_qc.html
index 6e7c4ba..f250bd7 100644
--- a/eda_qc.html
+++ b/eda_qc.html
@@ -486,7 +486,7 @@ Challenge
-
+
R
@@ -565,7 +565,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -791,7 +791,7 @@
Challenge
-
+
R
@@ -1063,7 +1063,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1349,7 +1349,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -1512,7 +1512,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
diff --git a/hca.html b/hca.html
index 18576b9..071b6cc 100644
--- a/hca.html
+++ b/hca.html
@@ -788,7 +788,7 @@
Exercise 1
-
+
R
@@ -821,7 +821,7 @@ Exercise 2
-
+
R
@@ -853,7 +853,7 @@ Exercise 3
-
+
R
@@ -894,7 +894,7 @@ Exercise 4
-
+
R
diff --git a/index.html b/index.html
index 24dd0f3..6dfcdaf 100644
--- a/index.html
+++ b/index.html
@@ -372,7 +372,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -398,7 +398,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN
website.
@@ -429,7 +429,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -453,7 +453,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN website.
- Select the
.pkg file for the latest R version
@@ -486,7 +486,7 @@
Install R using your package manager and RStudio
-
+
- Follow the instructions for your distribution from CRAN, they provide
information to get the most recent version of R for common
diff --git a/instructor/aio.html b/instructor/aio.html
index 4714d2d..f7a8835 100644
--- a/instructor/aio.html
+++ b/instructor/aio.html
@@ -1040,7 +1040,7 @@
Challenge
-
+
R
@@ -1120,7 +1120,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -1353,7 +1353,7 @@
Challenge
-
+
R
@@ -1629,7 +1629,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1924,7 +1924,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -2094,7 +2094,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
@@ -2380,7 +2380,7 @@
Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -2533,7 +2533,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -2826,7 +2826,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -3015,7 +3015,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -3024,231 +3026,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -3292,7 +3086,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -3441,7 +3235,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -3458,7 +3252,7 @@
Give me a hint
-
+
TODO
@@ -3487,7 +3281,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -3501,7 +3295,7 @@
Give me a hint
-
+
TODO
@@ -3531,7 +3325,7 @@ Exercise 3: Workflow
-
+
TODO
@@ -3603,7 +3397,7 @@ Key Points
-->
Content from Multi-sample analyses
-Last updated on 2024-09-06 |
+
Last updated on 2024-09-08 |
Edit this page
Estimated time: 45 minutes
@@ -3661,13 +3455,23 @@ Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.
We start by loading the data and doing a quick exploratory analysis,
essentially applying the normalization and visualization techniques that
-we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this
+time we’re selecting samples 5 to 10, not just 5 by itself.
R
library(MouseGastrulationData)
-sce <- WTChimeraData(samples=5:10, type = "processed")
+library(batchelor)
+library(edgeR)
+library(scater)
+library(ggplot2)
+library(scran)
+library(pheatmap)
+library(scuttle)
+
+sce <- WTChimeraData(samples = 5:10, type = "processed")
+
sce
@@ -3742,15 +3546,12 @@ OUTPUT<
R
-library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3758,23 +3559,24 @@ R
sce <- sce[,idx]
-We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps,
+and visualize them in a tSNE plot.
R
-# normalization
-sce <- logNormCounts(sce)
+sce <- logNormCounts(sce)
+
+dec <- modelGeneVar(sce, block = sce$sample)
-# identify highly variable genes
-dec <- modelGeneVar(sce, block=sce$sample)
chosen.hvgs <- dec$bio > 0
-# dimensionality reduction
sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000)
+
sce <- runTSNE(sce, dimred = "PCA")
sce$sample <- as.factor(sce$sample)
+
plotTSNE(sce, colour_by = "sample")
Challenge
-
+
Samples 5 and 6 were from the same “pool” of cells. Looking at the
documentation for the dataset under ?WTChimeraData we see
@@ -3822,7 +3624,7 @@
Show me the solution
first TSNE plot. If you were developing single-cell library preparation
protocols you might want to preserve this effect to understand how
variation in pools leads to variation in expression, but for now, given
-that we’re investigate other effects, we’ll want to remove this as
+that we’re investigating other effects, we’ll want to remove this as
undesired technical variation.
@@ -3831,29 +3633,30 @@ Show me the solution
Correcting batch effects
-We correct the effect of samples by aid of the
+
We “correct” the effect of samples with the
correctExperiment function in the batchelor
-package and using the sample colData column as
-batch.
+package and using the sample column as batch.
R
-library(batchelor)
-set.seed(10102)
-merged <- correctExperiments(sce,
- batch=sce$sample,
- subset.row=chosen.hvgs,
- PARAM=FastMnnParam(
- merge.order=list(
+set.seed(10102)
+
+merged <- correctExperiments(
+ sce,
+ batch = sce$sample,
+ subset.row = chosen.hvgs,
+ PARAM = FastMnnParam(
+ merge.order = list(
list(1,3,5), # WT (3 replicates)
list(2,4,6) # td-Tomato (3 replicates)
)
)
)
-merged <- runTSNE(merged, dimred="corrected")
-plotTSNE(merged, colour_by="batch")
+merged <- runTSNE(merged, dimred = "corrected")
+
+plotTSNE(merged, colour_by = "batch")
Once we removed the sample batch effect, we can proceed with the
Differential Expression Analysis.
@@ -3875,7 +3678,7 @@ Challenge
-
+
It’s important to have multiple samples within each experimental
group because it helps the batch effect correction algorithm distinguish
@@ -3898,7 +3701,7 @@
Show me the solution
Differential Expression
-In order to perform a Differential Expression Analysis, we need to
+
In order to perform a differential expression analysis, we need to
identify groups of cells across samples/conditions (depending on the
experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell
@@ -3926,9 +3729,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output
# corresponds to one unique combination of these two factors.
-library(scuttle)
-summed <- aggregateAcrossCells(merged,
- id=colData(merged)[,c("celltype.mapped", "sample")])
+
+summed <- aggregateAcrossCells(
+ merged,
+ id = colData(merged)[,c("celltype.mapped", "sample")]
+)
+
summed
R
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
y
@@ -4018,7 +3824,9 @@ R
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)
@@ -4034,8 +3842,10 @@ OUTPUT<
R
-keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)
@@ -4055,6 +3865,7 @@ R
y <- calcNormFactors(y)
+
y$samples
@@ -4095,7 +3906,7 @@ Rpar(mfrow = c(2,3))
for (i in seq_len(ncol(y))) {
- plotMD(y, column=i)
+ plotMD(y, column = i)
}
RR
-limma::plotMDS(cpm(y, log=TRUE),
- col=ifelse(y$samples$tomato, "red", "blue"))
+limma::plotMDS(cpm(y, log = TRUE),
+ col = ifelse(y$samples$tomato, "red", "blue"))
We then construct a design matrix by including both the pool and the
tomato as factors. This design indicates which samples belong to which
@@ -4153,6 +3964,7 @@
R
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)
@@ -4172,7 +3984,7 @@ RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized
-linear model for each gene. The robust=TRUE parameter
+linear model for each gene. The robust = TRUE parameter
avoids distortions from highly variable clusters. The QL method includes
an additional dispersion parameter, useful to handle the uncertainty and
variability of the per-gene variance, which is not well estimated by the
@@ -4182,7 +3994,8 @@
RR
-fit <- glmQLFit(y, design, robust=TRUE)
+fit <- glmQLFit(y, design, robust = TRUE)
+
summary(fit$var.prior)
@@ -4221,7 +4034,8 @@ RR
-res <- glmQLFTest(fit, coef=ncol(design))
+res <- glmQLFTest(fit, coef = ncol(design))
+
summary(decideTests(res))
@@ -4261,14 +4075,14 @@ OUTPUT<
R
-library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)
The returned object is a list of DataFrames each with
@@ -4280,6 +4094,7 @@
R
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]
Challenge
-
+
“logFC” stands for log fold-change. Rather than reporting e.g. a
5-fold increase, it’s better to report a logFC of log(5) = 1.61.
@@ -4336,26 +4151,29 @@
Show me the solution
With DA we look for differences in cluster abundance across
conditions (the tomato injection in our case), rather than differences
in gene expression.
-We first setup some code and variables for further analysis, like
-quantifying the number of cells per each cell type and fit a model to
-catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this
-time we start our analysis on the computed abundances and without
+
Our first steps are quantifying the number of cells per each cell
+type and fitting a model to catch differences between the injected cells
+and the background.
+The process is very similar differential expression modeling, but
+this time we start our analysis on the computed abundances and without
normalizing the data with TMM.
R
-library(edgeR)
-abundances <- table(merged$celltype.mapped, merged$sample)
+abundances <- table(merged$celltype.mapped, merged$sample)
+
abundances <- unclass(abundances)
-# Attaching some column metadata.
+
extra.info <- colData(merged)[match(colnames(abundances), merged$sample),]
-y.ab <- DGEList(abundances, samples=extra.info)
+
+y.ab <- DGEList(abundances, samples = extra.info)
design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples)
-y.ab <- estimateDisp(y.ab, design, trend="none")
-fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)
+
+y.ab <- estimateDisp(y.ab, design, trend = "none")
+
+fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)
Background on compositional effect
@@ -4401,6 +4219,7 @@ R
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
@@ -4413,11 +4232,11 @@ OUTPUT<
R
-y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))
@@ -4433,7 +4252,7 @@ OUTPUT<
R
-topTags(res2, n=10)
+topTags(res2, n = 10)
OUTPUT
@@ -4469,7 +4288,8 @@ Testing against a log-fold
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
@@ -4550,70 +4370,72 @@ OUTPUT<
[8] base
other attached packages:
- [1] edgeR_4.2.0 limma_3.60.2
- [3] batchelor_1.20.0 scran_1.32.0
- [5] scater_1.32.0 ggplot2_3.5.1
- [7] scuttle_1.14.0 MouseGastrulationData_1.18.0
- [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0
-[11] SummarizedExperiment_1.34.0 Biobase_2.64.0
-[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1
-[15] IRanges_2.38.0 S4Vectors_0.42.0
-[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0
-[19] matrixStats_1.3.0 BiocStyle_2.32.0
+ [1] pheatmap_1.0.12 scran_1.32.0
+ [3] scater_1.32.0 ggplot2_3.5.1
+ [5] scuttle_1.14.0 edgeR_4.2.0
+ [7] limma_3.60.2 batchelor_1.20.0
+ [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0
+[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0
+[13] Biobase_2.64.0 GenomicRanges_1.56.0
+[15] GenomeInfoDb_1.40.1 IRanges_2.38.0
+[17] S4Vectors_0.42.0 BiocGenerics_0.50.0
+[19] MatrixGenerics_1.16.0 matrixStats_1.3.0
+[21] BiocStyle_2.32.0
loaded via a namespace (and not attached):
- [1] DBI_1.2.3 formatR_1.14
- [3] gridExtra_2.3 rlang_1.1.3
- [5] magrittr_2.0.3 compiler_4.4.1
- [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0
- [9] png_0.1-8 vctrs_0.6.5
- [11] pkgconfig_2.0.3 crayon_1.5.2
- [13] fastmap_1.2.0 dbplyr_2.5.0
- [15] magick_2.8.3 XVector_0.44.0
- [17] labeling_0.4.3 utf8_1.2.4
- [19] rmarkdown_2.27 UCSC.utils_1.0.0
- [21] ggbeeswarm_0.7.2 purrr_1.0.2
- [23] bit_4.0.5 bluster_1.14.0
- [25] xfun_0.44 zlibbioc_1.50.0
- [27] cachem_1.1.0 beachmat_2.20.0
- [29] jsonlite_1.8.8 blob_1.2.4
- [31] highr_0.11 DelayedArray_0.30.1
- [33] BiocParallel_1.38.0 cluster_2.1.6
- [35] irlba_2.3.5.1 parallel_4.4.1
- [37] R6_2.5.1 Rcpp_1.0.12
- [39] knitr_1.47 splines_4.4.1
- [41] igraph_2.0.3 Matrix_1.7-0
- [43] tidyselect_1.2.1 viridis_0.6.5
- [45] abind_1.4-5 yaml_2.3.8
- [47] codetools_0.2-20 curl_5.2.1
- [49] lattice_0.22-6 tibble_3.2.1
- [51] withr_3.0.0 KEGGREST_1.44.0
- [53] BumpyMatrix_1.12.0 Rtsne_0.17
- [55] evaluate_0.23 BiocFileCache_2.12.0
- [57] ExperimentHub_2.12.0 Biostrings_2.72.1
- [59] pillar_1.9.0 BiocManager_1.30.23
- [61] filelock_1.0.3 renv_1.0.7
- [63] generics_0.1.3 BiocVersion_3.19.1
- [65] sparseMatrixStats_1.16.0 munsell_0.5.1
- [67] scales_1.3.0 glue_1.7.0
- [69] metapod_1.12.0 tools_4.4.1
- [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0
- [73] ScaledMatrix_1.12.0 locfit_1.5-9.9
- [75] cowplot_1.1.3 grid_4.4.1
- [77] AnnotationDbi_1.66.0 colorspace_2.1-0
- [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0
- [81] BiocSingular_1.20.0 vipor_0.4.7
- [83] cli_3.6.2 rsvd_1.0.5
- [85] rappdirs_0.3.3 fansi_1.0.6
- [87] viridisLite_0.4.2 S4Arrays_1.4.1
- [89] dplyr_1.1.4 ResidualMatrix_1.14.0
- [91] gtable_0.3.5 digest_0.6.35
- [93] dqrng_0.4.1 SparseArray_1.4.8
- [95] ggrepel_0.9.5 farver_2.1.2
- [97] rjson_0.2.21 memoise_2.0.1
- [99] htmltools_0.5.8.1 lifecycle_1.0.4
-[101] httr_1.4.7 statmod_1.5.0
-[103] mime_0.12 bit64_4.0.5
+ [1] RColorBrewer_1.1-3 jsonlite_1.8.8
+ [3] magrittr_2.0.3 ggbeeswarm_0.7.2
+ [5] magick_2.8.3 farver_2.1.2
+ [7] rmarkdown_2.27 zlibbioc_1.50.0
+ [9] vctrs_0.6.5 memoise_2.0.1
+ [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1
+ [13] S4Arrays_1.4.1 AnnotationHub_3.12.0
+ [15] curl_5.2.1 BiocNeighbors_1.22.0
+ [17] SparseArray_1.4.8 cachem_1.1.0
+ [19] ResidualMatrix_1.14.0 igraph_2.0.3
+ [21] mime_0.12 lifecycle_1.0.4
+ [23] pkgconfig_2.0.3 rsvd_1.0.5
+ [25] Matrix_1.7-0 R6_2.5.1
+ [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12
+ [29] digest_0.6.35 colorspace_2.1-0
+ [31] AnnotationDbi_1.66.0 dqrng_0.4.1
+ [33] irlba_2.3.5.1 ExperimentHub_2.12.0
+ [35] RSQLite_2.3.7 beachmat_2.20.0
+ [37] filelock_1.0.3 labeling_0.4.3
+ [39] fansi_1.0.6 httr_1.4.7
+ [41] abind_1.4-5 compiler_4.4.1
+ [43] bit64_4.0.5 withr_3.0.0
+ [45] BiocParallel_1.38.0 viridis_0.6.5
+ [47] DBI_1.2.3 highr_0.11
+ [49] rappdirs_0.3.3 DelayedArray_0.30.1
+ [51] rjson_0.2.21 bluster_1.14.0
+ [53] tools_4.4.1 vipor_0.4.7
+ [55] beeswarm_0.4.0 glue_1.7.0
+ [57] grid_4.4.1 Rtsne_0.17
+ [59] cluster_2.1.6 generics_0.1.3
+ [61] gtable_0.3.5 BiocSingular_1.20.0
+ [63] ScaledMatrix_1.12.0 metapod_1.12.0
+ [65] utf8_1.2.4 XVector_0.44.0
+ [67] ggrepel_0.9.5 BiocVersion_3.19.1
+ [69] pillar_1.9.0 BumpyMatrix_1.12.0
+ [71] splines_4.4.1 dplyr_1.1.4
+ [73] BiocFileCache_2.12.0 lattice_0.22-6
+ [75] renv_1.0.7 bit_4.0.5
+ [77] tidyselect_1.2.1 locfit_1.5-9.9
+ [79] Biostrings_2.72.1 knitr_1.47
+ [81] gridExtra_2.3 xfun_0.44
+ [83] statmod_1.5.0 UCSC.utils_1.0.0
+ [85] yaml_2.3.8 evaluate_0.23
+ [87] codetools_0.2-20 tibble_3.2.1
+ [89] BiocManager_1.30.23 cli_3.6.2
+ [91] munsell_0.5.1 Rcpp_1.0.12
+ [93] dbplyr_2.5.0 png_0.1-8
+ [95] parallel_4.4.1 blob_1.2.4
+ [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2
+ [99] scales_1.3.0 purrr_1.0.2
+[101] crayon_1.5.2 rlang_1.1.3
+[103] cowplot_1.1.3 KEGGREST_1.44.0
+[105] formatR_1.14
Exercises
@@ -4636,7 +4458,7 @@ Exercise 1: Replicates
-
+
Remember, you can subset SingleCellExperiments with logical indices,
just like a matrix. You can also access their column data with the
@@ -4650,7 +4472,7 @@
Give me a hint
-
+
R
@@ -4658,11 +4480,12 @@ R
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
@@ -4686,7 +4509,7 @@ Exercise 2:
-
+
You can just hand pheatmap() a matrix as its only
argument. It has a million options, but the defaults are usually pretty
@@ -4700,15 +4523,13 @@
Give me a hint
-
+
R
-library(pheatmap)
-
-pheatmap(y.ab$counts)
+pheatmap(y.ab$counts)
The top DA result was a decrease in ExE ectoderm in the tomato
condition, which you can sort of see, especially if you
@@ -6470,7 +6291,7 @@
Exercise 1
-
+
R
@@ -6503,7 +6324,7 @@ Exercise 2
-
+
R
@@ -6535,7 +6356,7 @@ Exercise 3
-
+
Whenever your code involves the generation of random numbers, it’s a good practice to set the random seed in R with @@ -1349,7 +1349,7 @@
Challenge
-R
@@ -1625,7 +1625,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1920,7 +1920,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -2090,7 +2090,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
@@ -2375,7 +2375,7 @@
Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -2528,7 +2528,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -2821,7 +2821,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -3010,7 +3010,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -3019,231 +3021,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -3287,7 +3081,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -3436,7 +3230,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -3453,7 +3247,7 @@
Give me a hint
-
+
TODO
@@ -3482,7 +3276,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -3496,7 +3290,7 @@
Give me a hint
-
+
TODO
@@ -3526,7 +3320,7 @@ Exercise 3: Workflow
-
+
TODO
@@ -3598,7 +3392,7 @@ Key Points
-->
Content from Multi-sample analyses
-Last updated on 2024-09-06 |
+
Last updated on 2024-09-08 |
Edit this page
@@ -3655,13 +3449,23 @@ Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.
We start by loading the data and doing a quick exploratory analysis,
essentially applying the normalization and visualization techniques that
-we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this
+time we’re selecting samples 5 to 10, not just 5 by itself.
R
library(MouseGastrulationData)
-sce <- WTChimeraData(samples=5:10, type = "processed")
+library(batchelor)
+library(edgeR)
+library(scater)
+library(ggplot2)
+library(scran)
+library(pheatmap)
+library(scuttle)
+
+sce <- WTChimeraData(samples = 5:10, type = "processed")
+
sce
@@ -3736,15 +3540,12 @@ OUTPUT<
R
-library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3752,23 +3553,24 @@ R
sce <- sce[,idx]
-We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps,
+and visualize them in a tSNE plot.
R
-# normalization
-sce <- logNormCounts(sce)
+sce <- logNormCounts(sce)
+
+dec <- modelGeneVar(sce, block = sce$sample)
-# identify highly variable genes
-dec <- modelGeneVar(sce, block=sce$sample)
chosen.hvgs <- dec$bio > 0
-# dimensionality reduction
sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000)
+
sce <- runTSNE(sce, dimred = "PCA")
sce$sample <- as.factor(sce$sample)
+
plotTSNE(sce, colour_by = "sample")
Challenge
-
+
Samples 5 and 6 were from the same “pool” of cells. Looking at the
documentation for the dataset under ?WTChimeraData we see
@@ -3816,7 +3618,7 @@
Show me the solution
first TSNE plot. If you were developing single-cell library preparation
protocols you might want to preserve this effect to understand how
variation in pools leads to variation in expression, but for now, given
-that we’re investigate other effects, we’ll want to remove this as
+that we’re investigating other effects, we’ll want to remove this as
undesired technical variation.
@@ -3825,29 +3627,30 @@ Show me the solution
Correcting batch effects
-We correct the effect of samples by aid of the
+
We “correct” the effect of samples with the
correctExperiment function in the batchelor
-package and using the sample colData column as
-batch.
+package and using the sample column as batch.
R
-library(batchelor)
-set.seed(10102)
-merged <- correctExperiments(sce,
- batch=sce$sample,
- subset.row=chosen.hvgs,
- PARAM=FastMnnParam(
- merge.order=list(
+set.seed(10102)
+
+merged <- correctExperiments(
+ sce,
+ batch = sce$sample,
+ subset.row = chosen.hvgs,
+ PARAM = FastMnnParam(
+ merge.order = list(
list(1,3,5), # WT (3 replicates)
list(2,4,6) # td-Tomato (3 replicates)
)
)
)
-merged <- runTSNE(merged, dimred="corrected")
-plotTSNE(merged, colour_by="batch")
+merged <- runTSNE(merged, dimred = "corrected")
+
+plotTSNE(merged, colour_by = "batch")
Once we removed the sample batch effect, we can proceed with the
Differential Expression Analysis.
@@ -3869,7 +3672,7 @@ Challenge
-
+
It’s important to have multiple samples within each experimental
group because it helps the batch effect correction algorithm distinguish
@@ -3892,7 +3695,7 @@
Show me the solution
Differential Expression
-In order to perform a Differential Expression Analysis, we need to
+
In order to perform a differential expression analysis, we need to
identify groups of cells across samples/conditions (depending on the
experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell
@@ -3920,9 +3723,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output
# corresponds to one unique combination of these two factors.
-library(scuttle)
-summed <- aggregateAcrossCells(merged,
- id=colData(merged)[,c("celltype.mapped", "sample")])
+
+summed <- aggregateAcrossCells(
+ merged,
+ id = colData(merged)[,c("celltype.mapped", "sample")]
+)
+
summed
R
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
y
@@ -4012,7 +3818,9 @@ R
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)
@@ -4028,8 +3836,10 @@ OUTPUT<
R
-keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)
@@ -4049,6 +3859,7 @@ R
y <- calcNormFactors(y)
+
y$samples
@@ -4089,7 +3900,7 @@ Rpar(mfrow = c(2,3))
for (i in seq_len(ncol(y))) {
- plotMD(y, column=i)
+ plotMD(y, column = i)
}
RR
-limma::plotMDS(cpm(y, log=TRUE),
- col=ifelse(y$samples$tomato, "red", "blue"))
+limma::plotMDS(cpm(y, log = TRUE),
+ col = ifelse(y$samples$tomato, "red", "blue"))
We then construct a design matrix by including both the pool and the
tomato as factors. This design indicates which samples belong to which
@@ -4147,6 +3958,7 @@
R
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)
@@ -4166,7 +3978,7 @@ RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized
-linear model for each gene. The robust=TRUE parameter
+linear model for each gene. The robust = TRUE parameter
avoids distortions from highly variable clusters. The QL method includes
an additional dispersion parameter, useful to handle the uncertainty and
variability of the per-gene variance, which is not well estimated by the
@@ -4176,7 +3988,8 @@
RR
-fit <- glmQLFit(y, design, robust=TRUE)
+fit <- glmQLFit(y, design, robust = TRUE)
+
summary(fit$var.prior)
@@ -4215,7 +4028,8 @@ RR
-res <- glmQLFTest(fit, coef=ncol(design))
+res <- glmQLFTest(fit, coef = ncol(design))
+
summary(decideTests(res))
@@ -4255,14 +4069,14 @@ OUTPUT<
R
-library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)
The returned object is a list of DataFrames each with
@@ -4274,6 +4088,7 @@
R
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]
Challenge
-
+
“logFC” stands for log fold-change. Rather than reporting e.g. a
5-fold increase, it’s better to report a logFC of log(5) = 1.61.
@@ -4330,26 +4145,29 @@
Show me the solution
With DA we look for differences in cluster abundance across
conditions (the tomato injection in our case), rather than differences
in gene expression.
-We first setup some code and variables for further analysis, like
-quantifying the number of cells per each cell type and fit a model to
-catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this
-time we start our analysis on the computed abundances and without
+
Our first steps are quantifying the number of cells per each cell
+type and fitting a model to catch differences between the injected cells
+and the background.
+The process is very similar differential expression modeling, but
+this time we start our analysis on the computed abundances and without
normalizing the data with TMM.
R
-library(edgeR)
-abundances <- table(merged$celltype.mapped, merged$sample)
+abundances <- table(merged$celltype.mapped, merged$sample)
+
abundances <- unclass(abundances)
-# Attaching some column metadata.
+
extra.info <- colData(merged)[match(colnames(abundances), merged$sample),]
-y.ab <- DGEList(abundances, samples=extra.info)
+
+y.ab <- DGEList(abundances, samples = extra.info)
design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples)
-y.ab <- estimateDisp(y.ab, design, trend="none")
-fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)
+
+y.ab <- estimateDisp(y.ab, design, trend = "none")
+
+fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)
Background on compositional effect
@@ -4395,6 +4213,7 @@ R
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
@@ -4407,11 +4226,11 @@ OUTPUT<
R
-y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))
@@ -4427,7 +4246,7 @@ OUTPUT<
R
-topTags(res2, n=10)
+topTags(res2, n = 10)
OUTPUT
@@ -4463,7 +4282,8 @@ Testing against a log-fold
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
@@ -4544,70 +4364,72 @@ OUTPUT<
[8] base
other attached packages:
- [1] edgeR_4.2.0 limma_3.60.2
- [3] batchelor_1.20.0 scran_1.32.0
- [5] scater_1.32.0 ggplot2_3.5.1
- [7] scuttle_1.14.0 MouseGastrulationData_1.18.0
- [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0
-[11] SummarizedExperiment_1.34.0 Biobase_2.64.0
-[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1
-[15] IRanges_2.38.0 S4Vectors_0.42.0
-[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0
-[19] matrixStats_1.3.0 BiocStyle_2.32.0
+ [1] pheatmap_1.0.12 scran_1.32.0
+ [3] scater_1.32.0 ggplot2_3.5.1
+ [5] scuttle_1.14.0 edgeR_4.2.0
+ [7] limma_3.60.2 batchelor_1.20.0
+ [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0
+[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0
+[13] Biobase_2.64.0 GenomicRanges_1.56.0
+[15] GenomeInfoDb_1.40.1 IRanges_2.38.0
+[17] S4Vectors_0.42.0 BiocGenerics_0.50.0
+[19] MatrixGenerics_1.16.0 matrixStats_1.3.0
+[21] BiocStyle_2.32.0
loaded via a namespace (and not attached):
- [1] DBI_1.2.3 formatR_1.14
- [3] gridExtra_2.3 rlang_1.1.3
- [5] magrittr_2.0.3 compiler_4.4.1
- [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0
- [9] png_0.1-8 vctrs_0.6.5
- [11] pkgconfig_2.0.3 crayon_1.5.2
- [13] fastmap_1.2.0 dbplyr_2.5.0
- [15] magick_2.8.3 XVector_0.44.0
- [17] labeling_0.4.3 utf8_1.2.4
- [19] rmarkdown_2.27 UCSC.utils_1.0.0
- [21] ggbeeswarm_0.7.2 purrr_1.0.2
- [23] bit_4.0.5 bluster_1.14.0
- [25] xfun_0.44 zlibbioc_1.50.0
- [27] cachem_1.1.0 beachmat_2.20.0
- [29] jsonlite_1.8.8 blob_1.2.4
- [31] highr_0.11 DelayedArray_0.30.1
- [33] BiocParallel_1.38.0 cluster_2.1.6
- [35] irlba_2.3.5.1 parallel_4.4.1
- [37] R6_2.5.1 Rcpp_1.0.12
- [39] knitr_1.47 splines_4.4.1
- [41] igraph_2.0.3 Matrix_1.7-0
- [43] tidyselect_1.2.1 viridis_0.6.5
- [45] abind_1.4-5 yaml_2.3.8
- [47] codetools_0.2-20 curl_5.2.1
- [49] lattice_0.22-6 tibble_3.2.1
- [51] withr_3.0.0 KEGGREST_1.44.0
- [53] BumpyMatrix_1.12.0 Rtsne_0.17
- [55] evaluate_0.23 BiocFileCache_2.12.0
- [57] ExperimentHub_2.12.0 Biostrings_2.72.1
- [59] pillar_1.9.0 BiocManager_1.30.23
- [61] filelock_1.0.3 renv_1.0.7
- [63] generics_0.1.3 BiocVersion_3.19.1
- [65] sparseMatrixStats_1.16.0 munsell_0.5.1
- [67] scales_1.3.0 glue_1.7.0
- [69] metapod_1.12.0 tools_4.4.1
- [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0
- [73] ScaledMatrix_1.12.0 locfit_1.5-9.9
- [75] cowplot_1.1.3 grid_4.4.1
- [77] AnnotationDbi_1.66.0 colorspace_2.1-0
- [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0
- [81] BiocSingular_1.20.0 vipor_0.4.7
- [83] cli_3.6.2 rsvd_1.0.5
- [85] rappdirs_0.3.3 fansi_1.0.6
- [87] viridisLite_0.4.2 S4Arrays_1.4.1
- [89] dplyr_1.1.4 ResidualMatrix_1.14.0
- [91] gtable_0.3.5 digest_0.6.35
- [93] dqrng_0.4.1 SparseArray_1.4.8
- [95] ggrepel_0.9.5 farver_2.1.2
- [97] rjson_0.2.21 memoise_2.0.1
- [99] htmltools_0.5.8.1 lifecycle_1.0.4
-[101] httr_1.4.7 statmod_1.5.0
-[103] mime_0.12 bit64_4.0.5
+ [1] RColorBrewer_1.1-3 jsonlite_1.8.8
+ [3] magrittr_2.0.3 ggbeeswarm_0.7.2
+ [5] magick_2.8.3 farver_2.1.2
+ [7] rmarkdown_2.27 zlibbioc_1.50.0
+ [9] vctrs_0.6.5 memoise_2.0.1
+ [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1
+ [13] S4Arrays_1.4.1 AnnotationHub_3.12.0
+ [15] curl_5.2.1 BiocNeighbors_1.22.0
+ [17] SparseArray_1.4.8 cachem_1.1.0
+ [19] ResidualMatrix_1.14.0 igraph_2.0.3
+ [21] mime_0.12 lifecycle_1.0.4
+ [23] pkgconfig_2.0.3 rsvd_1.0.5
+ [25] Matrix_1.7-0 R6_2.5.1
+ [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12
+ [29] digest_0.6.35 colorspace_2.1-0
+ [31] AnnotationDbi_1.66.0 dqrng_0.4.1
+ [33] irlba_2.3.5.1 ExperimentHub_2.12.0
+ [35] RSQLite_2.3.7 beachmat_2.20.0
+ [37] filelock_1.0.3 labeling_0.4.3
+ [39] fansi_1.0.6 httr_1.4.7
+ [41] abind_1.4-5 compiler_4.4.1
+ [43] bit64_4.0.5 withr_3.0.0
+ [45] BiocParallel_1.38.0 viridis_0.6.5
+ [47] DBI_1.2.3 highr_0.11
+ [49] rappdirs_0.3.3 DelayedArray_0.30.1
+ [51] rjson_0.2.21 bluster_1.14.0
+ [53] tools_4.4.1 vipor_0.4.7
+ [55] beeswarm_0.4.0 glue_1.7.0
+ [57] grid_4.4.1 Rtsne_0.17
+ [59] cluster_2.1.6 generics_0.1.3
+ [61] gtable_0.3.5 BiocSingular_1.20.0
+ [63] ScaledMatrix_1.12.0 metapod_1.12.0
+ [65] utf8_1.2.4 XVector_0.44.0
+ [67] ggrepel_0.9.5 BiocVersion_3.19.1
+ [69] pillar_1.9.0 BumpyMatrix_1.12.0
+ [71] splines_4.4.1 dplyr_1.1.4
+ [73] BiocFileCache_2.12.0 lattice_0.22-6
+ [75] renv_1.0.7 bit_4.0.5
+ [77] tidyselect_1.2.1 locfit_1.5-9.9
+ [79] Biostrings_2.72.1 knitr_1.47
+ [81] gridExtra_2.3 xfun_0.44
+ [83] statmod_1.5.0 UCSC.utils_1.0.0
+ [85] yaml_2.3.8 evaluate_0.23
+ [87] codetools_0.2-20 tibble_3.2.1
+ [89] BiocManager_1.30.23 cli_3.6.2
+ [91] munsell_0.5.1 Rcpp_1.0.12
+ [93] dbplyr_2.5.0 png_0.1-8
+ [95] parallel_4.4.1 blob_1.2.4
+ [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2
+ [99] scales_1.3.0 purrr_1.0.2
+[101] crayon_1.5.2 rlang_1.1.3
+[103] cowplot_1.1.3 KEGGREST_1.44.0
+[105] formatR_1.14
Exercises
@@ -4630,7 +4452,7 @@ Exercise 1: Replicates
-
+
Remember, you can subset SingleCellExperiments with logical indices,
just like a matrix. You can also access their column data with the
@@ -4644,7 +4466,7 @@
Give me a hint
-
+
R
@@ -4652,11 +4474,12 @@ R
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
@@ -4680,7 +4503,7 @@ Exercise 2:
-
+
You can just hand pheatmap() a matrix as its only
argument. It has a million options, but the defaults are usually pretty
@@ -4694,15 +4517,13 @@
Give me a hint
-
+
R
-library(pheatmap)
-
-pheatmap(y.ab$counts)
+pheatmap(y.ab$counts)
The top DA result was a decrease in ExE ectoderm in the tomato
condition, which you can sort of see, especially if you
@@ -6462,7 +6283,7 @@
Exercise 1
-
+
R
@@ -6495,7 +6316,7 @@ Exercise 2
-
+
R
@@ -6527,7 +6348,7 @@ Exercise 3
-
+
R
@@ -6568,7 +6389,7 @@ Exercise 4
-
+
R
diff --git a/cell_type_annotation.html b/cell_type_annotation.html
index 067a8e8..a423c5a 100644
--- a/cell_type_annotation.html
+++ b/cell_type_annotation.html
@@ -515,7 +515,7 @@ Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -666,7 +666,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -956,7 +956,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -1144,7 +1144,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -1153,231 +1155,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -1421,7 +1215,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -1566,7 +1360,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -1583,7 +1377,7 @@
Give me a hint
-
+
TODO
@@ -1612,7 +1406,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -1626,7 +1420,7 @@
Give me a hint
-
+
TODO
@@ -1656,7 +1450,7 @@ Exercise 3: Workflow
-
+
TODO
diff --git a/eda_qc.html b/eda_qc.html
index 6e7c4ba..f250bd7 100644
--- a/eda_qc.html
+++ b/eda_qc.html
@@ -486,7 +486,7 @@ Challenge
-
+
R
@@ -565,7 +565,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -791,7 +791,7 @@
Challenge
-
+
R
@@ -1063,7 +1063,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1349,7 +1349,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -1512,7 +1512,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
diff --git a/hca.html b/hca.html
index 18576b9..071b6cc 100644
--- a/hca.html
+++ b/hca.html
@@ -788,7 +788,7 @@
Exercise 1
-
+
R
@@ -821,7 +821,7 @@ Exercise 2
-
+
R
@@ -853,7 +853,7 @@ Exercise 3
-
+
R
@@ -894,7 +894,7 @@ Exercise 4
-
+
R
diff --git a/index.html b/index.html
index 24dd0f3..6dfcdaf 100644
--- a/index.html
+++ b/index.html
@@ -372,7 +372,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -398,7 +398,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN
website.
@@ -429,7 +429,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -453,7 +453,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN website.
- Select the
.pkg file for the latest R version
@@ -486,7 +486,7 @@
Install R using your package manager and RStudio
-
+
- Follow the instructions for your distribution from CRAN, they provide
information to get the most recent version of R for common
diff --git a/instructor/aio.html b/instructor/aio.html
index 4714d2d..f7a8835 100644
--- a/instructor/aio.html
+++ b/instructor/aio.html
@@ -1040,7 +1040,7 @@
Challenge
-
+
R
@@ -1120,7 +1120,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -1353,7 +1353,7 @@
Challenge
-
+
R
@@ -1629,7 +1629,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1924,7 +1924,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -2094,7 +2094,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
@@ -2380,7 +2380,7 @@
Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -2533,7 +2533,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -2826,7 +2826,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -3015,7 +3015,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -3024,231 +3026,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -3292,7 +3086,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -3441,7 +3235,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -3458,7 +3252,7 @@
Give me a hint
-
+
TODO
@@ -3487,7 +3281,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -3501,7 +3295,7 @@
Give me a hint
-
+
TODO
@@ -3531,7 +3325,7 @@ Exercise 3: Workflow
-
+
TODO
@@ -3603,7 +3397,7 @@ Key Points
-->
Content from Multi-sample analyses
-Last updated on 2024-09-06 |
+
Last updated on 2024-09-08 |
Edit this page
Estimated time: 45 minutes
@@ -3661,13 +3455,23 @@ Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.
We start by loading the data and doing a quick exploratory analysis,
essentially applying the normalization and visualization techniques that
-we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this
+time we’re selecting samples 5 to 10, not just 5 by itself.
R
library(MouseGastrulationData)
-sce <- WTChimeraData(samples=5:10, type = "processed")
+library(batchelor)
+library(edgeR)
+library(scater)
+library(ggplot2)
+library(scran)
+library(pheatmap)
+library(scuttle)
+
+sce <- WTChimeraData(samples = 5:10, type = "processed")
+
sce
@@ -3742,15 +3546,12 @@ OUTPUT<
R
-library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3758,23 +3559,24 @@ R
sce <- sce[,idx]
-We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps,
+and visualize them in a tSNE plot.
R
-# normalization
-sce <- logNormCounts(sce)
+sce <- logNormCounts(sce)
+
+dec <- modelGeneVar(sce, block = sce$sample)
-# identify highly variable genes
-dec <- modelGeneVar(sce, block=sce$sample)
chosen.hvgs <- dec$bio > 0
-# dimensionality reduction
sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000)
+
sce <- runTSNE(sce, dimred = "PCA")
sce$sample <- as.factor(sce$sample)
+
plotTSNE(sce, colour_by = "sample")
Challenge
-
+
Samples 5 and 6 were from the same “pool” of cells. Looking at the
documentation for the dataset under ?WTChimeraData we see
@@ -3822,7 +3624,7 @@
Show me the solution
first TSNE plot. If you were developing single-cell library preparation
protocols you might want to preserve this effect to understand how
variation in pools leads to variation in expression, but for now, given
-that we’re investigate other effects, we’ll want to remove this as
+that we’re investigating other effects, we’ll want to remove this as
undesired technical variation.
@@ -3831,29 +3633,30 @@ Show me the solution
Correcting batch effects
-We correct the effect of samples by aid of the
+
We “correct” the effect of samples with the
correctExperiment function in the batchelor
-package and using the sample colData column as
-batch.
+package and using the sample column as batch.
R
-library(batchelor)
-set.seed(10102)
-merged <- correctExperiments(sce,
- batch=sce$sample,
- subset.row=chosen.hvgs,
- PARAM=FastMnnParam(
- merge.order=list(
+set.seed(10102)
+
+merged <- correctExperiments(
+ sce,
+ batch = sce$sample,
+ subset.row = chosen.hvgs,
+ PARAM = FastMnnParam(
+ merge.order = list(
list(1,3,5), # WT (3 replicates)
list(2,4,6) # td-Tomato (3 replicates)
)
)
)
-merged <- runTSNE(merged, dimred="corrected")
-plotTSNE(merged, colour_by="batch")
+merged <- runTSNE(merged, dimred = "corrected")
+
+plotTSNE(merged, colour_by = "batch")
Once we removed the sample batch effect, we can proceed with the
Differential Expression Analysis.
@@ -3875,7 +3678,7 @@ Challenge
-
+
It’s important to have multiple samples within each experimental
group because it helps the batch effect correction algorithm distinguish
@@ -3898,7 +3701,7 @@
Show me the solution
Differential Expression
-In order to perform a Differential Expression Analysis, we need to
+
In order to perform a differential expression analysis, we need to
identify groups of cells across samples/conditions (depending on the
experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell
@@ -3926,9 +3729,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output
# corresponds to one unique combination of these two factors.
-library(scuttle)
-summed <- aggregateAcrossCells(merged,
- id=colData(merged)[,c("celltype.mapped", "sample")])
+
+summed <- aggregateAcrossCells(
+ merged,
+ id = colData(merged)[,c("celltype.mapped", "sample")]
+)
+
summed
R
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
y
@@ -4018,7 +3824,9 @@ R
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)
@@ -4034,8 +3842,10 @@ OUTPUT<
R
-keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)
@@ -4055,6 +3865,7 @@ R
y <- calcNormFactors(y)
+
y$samples
@@ -4095,7 +3906,7 @@ Rpar(mfrow = c(2,3))
for (i in seq_len(ncol(y))) {
- plotMD(y, column=i)
+ plotMD(y, column = i)
}
RR
-limma::plotMDS(cpm(y, log=TRUE),
- col=ifelse(y$samples$tomato, "red", "blue"))
+limma::plotMDS(cpm(y, log = TRUE),
+ col = ifelse(y$samples$tomato, "red", "blue"))
We then construct a design matrix by including both the pool and the
tomato as factors. This design indicates which samples belong to which
@@ -4153,6 +3964,7 @@
R
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)
@@ -4172,7 +3984,7 @@ RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized
-linear model for each gene. The robust=TRUE parameter
+linear model for each gene. The robust = TRUE parameter
avoids distortions from highly variable clusters. The QL method includes
an additional dispersion parameter, useful to handle the uncertainty and
variability of the per-gene variance, which is not well estimated by the
@@ -4182,7 +3994,8 @@
RR
-fit <- glmQLFit(y, design, robust=TRUE)
+fit <- glmQLFit(y, design, robust = TRUE)
+
summary(fit$var.prior)
@@ -4221,7 +4034,8 @@ RR
-res <- glmQLFTest(fit, coef=ncol(design))
+res <- glmQLFTest(fit, coef = ncol(design))
+
summary(decideTests(res))
@@ -4261,14 +4075,14 @@ OUTPUT<
R
-library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)
The returned object is a list of DataFrames each with
@@ -4280,6 +4094,7 @@
R
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]
Challenge
-
+
“logFC” stands for log fold-change. Rather than reporting e.g. a
5-fold increase, it’s better to report a logFC of log(5) = 1.61.
@@ -4336,26 +4151,29 @@
Show me the solution
With DA we look for differences in cluster abundance across
conditions (the tomato injection in our case), rather than differences
in gene expression.
-We first setup some code and variables for further analysis, like
-quantifying the number of cells per each cell type and fit a model to
-catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this
-time we start our analysis on the computed abundances and without
+
Our first steps are quantifying the number of cells per each cell
+type and fitting a model to catch differences between the injected cells
+and the background.
+The process is very similar differential expression modeling, but
+this time we start our analysis on the computed abundances and without
normalizing the data with TMM.
R
-library(edgeR)
-abundances <- table(merged$celltype.mapped, merged$sample)
+abundances <- table(merged$celltype.mapped, merged$sample)
+
abundances <- unclass(abundances)
-# Attaching some column metadata.
+
extra.info <- colData(merged)[match(colnames(abundances), merged$sample),]
-y.ab <- DGEList(abundances, samples=extra.info)
+
+y.ab <- DGEList(abundances, samples = extra.info)
design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples)
-y.ab <- estimateDisp(y.ab, design, trend="none")
-fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)
+
+y.ab <- estimateDisp(y.ab, design, trend = "none")
+
+fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)
Background on compositional effect
@@ -4401,6 +4219,7 @@ R
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
@@ -4413,11 +4232,11 @@ OUTPUT<
R
-y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))
@@ -4433,7 +4252,7 @@ OUTPUT<
R
-topTags(res2, n=10)
+topTags(res2, n = 10)
OUTPUT
@@ -4469,7 +4288,8 @@ Testing against a log-fold
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
@@ -4550,70 +4370,72 @@ OUTPUT<
[8] base
other attached packages:
- [1] edgeR_4.2.0 limma_3.60.2
- [3] batchelor_1.20.0 scran_1.32.0
- [5] scater_1.32.0 ggplot2_3.5.1
- [7] scuttle_1.14.0 MouseGastrulationData_1.18.0
- [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0
-[11] SummarizedExperiment_1.34.0 Biobase_2.64.0
-[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1
-[15] IRanges_2.38.0 S4Vectors_0.42.0
-[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0
-[19] matrixStats_1.3.0 BiocStyle_2.32.0
+ [1] pheatmap_1.0.12 scran_1.32.0
+ [3] scater_1.32.0 ggplot2_3.5.1
+ [5] scuttle_1.14.0 edgeR_4.2.0
+ [7] limma_3.60.2 batchelor_1.20.0
+ [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0
+[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0
+[13] Biobase_2.64.0 GenomicRanges_1.56.0
+[15] GenomeInfoDb_1.40.1 IRanges_2.38.0
+[17] S4Vectors_0.42.0 BiocGenerics_0.50.0
+[19] MatrixGenerics_1.16.0 matrixStats_1.3.0
+[21] BiocStyle_2.32.0
loaded via a namespace (and not attached):
- [1] DBI_1.2.3 formatR_1.14
- [3] gridExtra_2.3 rlang_1.1.3
- [5] magrittr_2.0.3 compiler_4.4.1
- [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0
- [9] png_0.1-8 vctrs_0.6.5
- [11] pkgconfig_2.0.3 crayon_1.5.2
- [13] fastmap_1.2.0 dbplyr_2.5.0
- [15] magick_2.8.3 XVector_0.44.0
- [17] labeling_0.4.3 utf8_1.2.4
- [19] rmarkdown_2.27 UCSC.utils_1.0.0
- [21] ggbeeswarm_0.7.2 purrr_1.0.2
- [23] bit_4.0.5 bluster_1.14.0
- [25] xfun_0.44 zlibbioc_1.50.0
- [27] cachem_1.1.0 beachmat_2.20.0
- [29] jsonlite_1.8.8 blob_1.2.4
- [31] highr_0.11 DelayedArray_0.30.1
- [33] BiocParallel_1.38.0 cluster_2.1.6
- [35] irlba_2.3.5.1 parallel_4.4.1
- [37] R6_2.5.1 Rcpp_1.0.12
- [39] knitr_1.47 splines_4.4.1
- [41] igraph_2.0.3 Matrix_1.7-0
- [43] tidyselect_1.2.1 viridis_0.6.5
- [45] abind_1.4-5 yaml_2.3.8
- [47] codetools_0.2-20 curl_5.2.1
- [49] lattice_0.22-6 tibble_3.2.1
- [51] withr_3.0.0 KEGGREST_1.44.0
- [53] BumpyMatrix_1.12.0 Rtsne_0.17
- [55] evaluate_0.23 BiocFileCache_2.12.0
- [57] ExperimentHub_2.12.0 Biostrings_2.72.1
- [59] pillar_1.9.0 BiocManager_1.30.23
- [61] filelock_1.0.3 renv_1.0.7
- [63] generics_0.1.3 BiocVersion_3.19.1
- [65] sparseMatrixStats_1.16.0 munsell_0.5.1
- [67] scales_1.3.0 glue_1.7.0
- [69] metapod_1.12.0 tools_4.4.1
- [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0
- [73] ScaledMatrix_1.12.0 locfit_1.5-9.9
- [75] cowplot_1.1.3 grid_4.4.1
- [77] AnnotationDbi_1.66.0 colorspace_2.1-0
- [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0
- [81] BiocSingular_1.20.0 vipor_0.4.7
- [83] cli_3.6.2 rsvd_1.0.5
- [85] rappdirs_0.3.3 fansi_1.0.6
- [87] viridisLite_0.4.2 S4Arrays_1.4.1
- [89] dplyr_1.1.4 ResidualMatrix_1.14.0
- [91] gtable_0.3.5 digest_0.6.35
- [93] dqrng_0.4.1 SparseArray_1.4.8
- [95] ggrepel_0.9.5 farver_2.1.2
- [97] rjson_0.2.21 memoise_2.0.1
- [99] htmltools_0.5.8.1 lifecycle_1.0.4
-[101] httr_1.4.7 statmod_1.5.0
-[103] mime_0.12 bit64_4.0.5
+ [1] RColorBrewer_1.1-3 jsonlite_1.8.8
+ [3] magrittr_2.0.3 ggbeeswarm_0.7.2
+ [5] magick_2.8.3 farver_2.1.2
+ [7] rmarkdown_2.27 zlibbioc_1.50.0
+ [9] vctrs_0.6.5 memoise_2.0.1
+ [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1
+ [13] S4Arrays_1.4.1 AnnotationHub_3.12.0
+ [15] curl_5.2.1 BiocNeighbors_1.22.0
+ [17] SparseArray_1.4.8 cachem_1.1.0
+ [19] ResidualMatrix_1.14.0 igraph_2.0.3
+ [21] mime_0.12 lifecycle_1.0.4
+ [23] pkgconfig_2.0.3 rsvd_1.0.5
+ [25] Matrix_1.7-0 R6_2.5.1
+ [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12
+ [29] digest_0.6.35 colorspace_2.1-0
+ [31] AnnotationDbi_1.66.0 dqrng_0.4.1
+ [33] irlba_2.3.5.1 ExperimentHub_2.12.0
+ [35] RSQLite_2.3.7 beachmat_2.20.0
+ [37] filelock_1.0.3 labeling_0.4.3
+ [39] fansi_1.0.6 httr_1.4.7
+ [41] abind_1.4-5 compiler_4.4.1
+ [43] bit64_4.0.5 withr_3.0.0
+ [45] BiocParallel_1.38.0 viridis_0.6.5
+ [47] DBI_1.2.3 highr_0.11
+ [49] rappdirs_0.3.3 DelayedArray_0.30.1
+ [51] rjson_0.2.21 bluster_1.14.0
+ [53] tools_4.4.1 vipor_0.4.7
+ [55] beeswarm_0.4.0 glue_1.7.0
+ [57] grid_4.4.1 Rtsne_0.17
+ [59] cluster_2.1.6 generics_0.1.3
+ [61] gtable_0.3.5 BiocSingular_1.20.0
+ [63] ScaledMatrix_1.12.0 metapod_1.12.0
+ [65] utf8_1.2.4 XVector_0.44.0
+ [67] ggrepel_0.9.5 BiocVersion_3.19.1
+ [69] pillar_1.9.0 BumpyMatrix_1.12.0
+ [71] splines_4.4.1 dplyr_1.1.4
+ [73] BiocFileCache_2.12.0 lattice_0.22-6
+ [75] renv_1.0.7 bit_4.0.5
+ [77] tidyselect_1.2.1 locfit_1.5-9.9
+ [79] Biostrings_2.72.1 knitr_1.47
+ [81] gridExtra_2.3 xfun_0.44
+ [83] statmod_1.5.0 UCSC.utils_1.0.0
+ [85] yaml_2.3.8 evaluate_0.23
+ [87] codetools_0.2-20 tibble_3.2.1
+ [89] BiocManager_1.30.23 cli_3.6.2
+ [91] munsell_0.5.1 Rcpp_1.0.12
+ [93] dbplyr_2.5.0 png_0.1-8
+ [95] parallel_4.4.1 blob_1.2.4
+ [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2
+ [99] scales_1.3.0 purrr_1.0.2
+[101] crayon_1.5.2 rlang_1.1.3
+[103] cowplot_1.1.3 KEGGREST_1.44.0
+[105] formatR_1.14
Exercises
@@ -4636,7 +4458,7 @@ Exercise 1: Replicates
-
+
Remember, you can subset SingleCellExperiments with logical indices,
just like a matrix. You can also access their column data with the
@@ -4650,7 +4472,7 @@
Give me a hint
-
+
R
@@ -4658,11 +4480,12 @@ R
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
@@ -4686,7 +4509,7 @@ Exercise 2:
-
+
You can just hand pheatmap() a matrix as its only
argument. It has a million options, but the defaults are usually pretty
@@ -4700,15 +4523,13 @@
Give me a hint
-
+
R
-library(pheatmap)
-
-pheatmap(y.ab$counts)
+pheatmap(y.ab$counts)
The top DA result was a decrease in ExE ectoderm in the tomato
condition, which you can sort of see, especially if you
@@ -6470,7 +6291,7 @@
Exercise 1
-
+
R
@@ -6503,7 +6324,7 @@ Exercise 2
-
+
R
@@ -6535,7 +6356,7 @@ Exercise 3
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic or synthetic source at a known concentration. This provides a known @@ -1920,7 +1920,7 @@
Challenge
-Mathematically, this would require the data to fall on a two-dimensional plane (for linear methods like PCA) or a smooth 2D @@ -2090,7 +2090,7 @@
Exercise 2: Dimensionality Reduction
-First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
@@ -2375,7 +2375,7 @@
Challenge
-We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -2528,7 +2528,7 @@
Challenge
-One important reason why is because averages over all other clusters can be sensitive to the cell type composition. If a rare cell type shows @@ -2821,7 +2821,7 @@
Challenge
-Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -3010,7 +3010,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -3019,231 +3021,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
R
@@ -3019,231 +3021,23 @@R tab <- table(new.labels, sce$celltype.mapped) -tab +tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20As a diagnostic measure, we examine the distribution of AUCs across cells for each label. In heterogeneous populations, the distribution for @@ -3287,7 +3081,7 @@
Challenge
-The example that jumps out most strongly to the eye is ExE endoderm, which doesn’t show clear separate modes. Simultaneously, Endothelium @@ -3436,7 +3230,7 @@
Exercise 1: Clustering
-The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -3453,7 +3247,7 @@
Give me a hint
-TODO
Exercise 2: Cluster annotation
-Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -3496,7 +3290,7 @@
Give me a hint
-TODO
Exercise 3: Workflow
-TODO
Key Points
-->Content from Multi-sample analyses
-
Last updated on 2024-09-06 | +
Last updated on 2024-09-08 | Edit this page
Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.We start by loading the data and doing a quick exploratory analysis, essentially applying the normalization and visualization techniques that -we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this +time we’re selecting samples 5 to 10, not just 5 by itself.R
library(MouseGastrulationData)
-sce <- WTChimeraData(samples=5:10, type = "processed")
+library(batchelor)
+library(edgeR)
+library(scater)
+library(ggplot2)
+library(scran)
+library(pheatmap)
+library(scuttle)
+
+sce <- WTChimeraData(samples = 5:10, type = "processed")
+
sce
OUTPUT<
R
-library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3752,23 +3553,24 @@ R
sce <- sce[,idx]
library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3752,23 +3553,24 @@ R
sce <- sce[,idx]
We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps, +and visualize them in a tSNE plot.
R
-# normalization
-sce <- logNormCounts(sce)
+sce <- logNormCounts(sce)
+
+dec <- modelGeneVar(sce, block = sce$sample)
-# identify highly variable genes
-dec <- modelGeneVar(sce, block=sce$sample)
chosen.hvgs <- dec$bio > 0
-# dimensionality reduction
sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000)
+
sce <- runTSNE(sce, dimred = "PCA")
sce$sample <- as.factor(sce$sample)
+
plotTSNE(sce, colour_by = "sample")
Challenge
-Samples 5 and 6 were from the same “pool” of cells. Looking at the
documentation for the dataset under ?WTChimeraData we see
@@ -3816,7 +3618,7 @@
Show me the solution
first TSNE plot. If you were developing single-cell library preparation protocols you might want to preserve this effect to understand how variation in pools leads to variation in expression, but for now, given -that we’re investigate other effects, we’ll want to remove this as +that we’re investigating other effects, we’ll want to remove this as undesired technical variation.Show me the solution
Correcting batch effects
-
We correct the effect of samples by aid of the +
We “correct” the effect of samples with the
correctExperiment function in the batchelor
-package and using the sample colData column as
-batch.
sample column as batch.
R
-library(batchelor)
-set.seed(10102)
-merged <- correctExperiments(sce,
- batch=sce$sample,
- subset.row=chosen.hvgs,
- PARAM=FastMnnParam(
- merge.order=list(
+set.seed(10102)
+
+merged <- correctExperiments(
+ sce,
+ batch = sce$sample,
+ subset.row = chosen.hvgs,
+ PARAM = FastMnnParam(
+ merge.order = list(
list(1,3,5), # WT (3 replicates)
list(2,4,6) # td-Tomato (3 replicates)
)
)
)
-merged <- runTSNE(merged, dimred="corrected")
-plotTSNE(merged, colour_by="batch")
+merged <- runTSNE(merged, dimred = "corrected")
+
+plotTSNE(merged, colour_by = "batch")
Once we removed the sample batch effect, we can proceed with the Differential Expression Analysis.
@@ -3869,7 +3672,7 @@Challenge
-It’s important to have multiple samples within each experimental group because it helps the batch effect correction algorithm distinguish @@ -3892,7 +3695,7 @@
Show me the solution
Differential Expression
-
In order to perform a Differential Expression Analysis, we need to +
In order to perform a differential expression analysis, we need to identify groups of cells across samples/conditions (depending on the experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell @@ -3920,9 +3723,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output
# corresponds to one unique combination of these two factors.
-library(scuttle)
-summed <- aggregateAcrossCells(merged,
- id=colData(merged)[,c("celltype.mapped", "sample")])
+
+summed <- aggregateAcrossCells(
+ merged,
+ id = colData(merged)[,c("celltype.mapped", "sample")]
+)
+
summed
R
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
y
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
yR
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)OUTPUT<
R
-keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)
keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)R
y <- calcNormFactors(y)
+
y$samples
y <- calcNormFactors(y)
+
y$samplesRpar(mfrow = c(2,3)) for (i in seq_len(ncol(y))) { - plotMD(y, column=i) + plotMD(y, column = i) }
RR
-limma::plotMDS(cpm(y, log=TRUE),
- col=ifelse(y$samples$tomato, "red", "blue"))
+limma::plotMDS(cpm(y, log = TRUE),
+ col = ifelse(y$samples$tomato, "red", "blue"))
We then construct a design matrix by including both the pool and the tomato as factors. This design indicates which samples belong to which @@ -4147,6 +3958,7 @@
R
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized
-linear model for each gene. The robust=TRUE parameter
+linear model for each gene. The robust = TRUE parameter
avoids distortions from highly variable clusters. The QL method includes
an additional dispersion parameter, useful to handle the uncertainty and
variability of the per-gene variance, which is not well estimated by the
@@ -4176,7 +3988,8 @@
RR
-fit <- glmQLFit(y, design, robust=TRUE)
+fit <- glmQLFit(y, design, robust = TRUE)
+
summary(fit$var.prior)
RR
-res <- glmQLFTest(fit, coef=ncol(design))
+res <- glmQLFTest(fit, coef = ncol(design))
+
summary(decideTests(res))
OUTPUT<
R
-library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)
library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)The returned object is a list of DataFrames each with
@@ -4274,6 +4088,7 @@
R
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]Challenge
-“logFC” stands for log fold-change. Rather than reporting e.g. a 5-fold increase, it’s better to report a logFC of log(5) = 1.61. @@ -4330,26 +4145,29 @@
Show me the solution
With DA we look for differences in cluster abundance across conditions (the tomato injection in our case), rather than differences in gene expression.
-We first setup some code and variables for further analysis, like -quantifying the number of cells per each cell type and fit a model to -catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this -time we start our analysis on the computed abundances and without +
Our first steps are quantifying the number of cells per each cell +type and fitting a model to catch differences between the injected cells +and the background.
+The process is very similar differential expression modeling, but +this time we start our analysis on the computed abundances and without normalizing the data with TMM.
R
-library(edgeR)
-abundances <- table(merged$celltype.mapped, merged$sample)
+abundances <- table(merged$celltype.mapped, merged$sample)
+
abundances <- unclass(abundances)
-# Attaching some column metadata.
+
extra.info <- colData(merged)[match(colnames(abundances), merged$sample),]
-y.ab <- DGEList(abundances, samples=extra.info)
+
+y.ab <- DGEList(abundances, samples = extra.info)
design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples)
-y.ab <- estimateDisp(y.ab, design, trend="none")
-fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)
+
+y.ab <- estimateDisp(y.ab, design, trend = "none")
+
+fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)
Background on compositional effect
@@ -4395,6 +4213,7 @@ R
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
OUTPUT<
R
-y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))
y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))OUTPUT<
R
-topTags(res2, n=10)
+topTags(res2, n = 10)
topTags(res2, n=10)OUTPUT
@@ -4463,7 +4282,8 @@ Testing against a log-fold
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
OUTPUT< [8] base other attached packages: - [1] edgeR_4.2.0 limma_3.60.2 - [3] batchelor_1.20.0 scran_1.32.0 - [5] scater_1.32.0 ggplot2_3.5.1 - [7] scuttle_1.14.0 MouseGastrulationData_1.18.0 - [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0 -[11] SummarizedExperiment_1.34.0 Biobase_2.64.0 -[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1 -[15] IRanges_2.38.0 S4Vectors_0.42.0 -[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0 -[19] matrixStats_1.3.0 BiocStyle_2.32.0 + [1] pheatmap_1.0.12 scran_1.32.0 + [3] scater_1.32.0 ggplot2_3.5.1 + [5] scuttle_1.14.0 edgeR_4.2.0 + [7] limma_3.60.2 batchelor_1.20.0 + [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0 +[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0 +[13] Biobase_2.64.0 GenomicRanges_1.56.0 +[15] GenomeInfoDb_1.40.1 IRanges_2.38.0 +[17] S4Vectors_0.42.0 BiocGenerics_0.50.0 +[19] MatrixGenerics_1.16.0 matrixStats_1.3.0 +[21] BiocStyle_2.32.0 loaded via a namespace (and not attached): - [1] DBI_1.2.3 formatR_1.14 - [3] gridExtra_2.3 rlang_1.1.3 - [5] magrittr_2.0.3 compiler_4.4.1 - [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0 - [9] png_0.1-8 vctrs_0.6.5 - [11] pkgconfig_2.0.3 crayon_1.5.2 - [13] fastmap_1.2.0 dbplyr_2.5.0 - [15] magick_2.8.3 XVector_0.44.0 - [17] labeling_0.4.3 utf8_1.2.4 - [19] rmarkdown_2.27 UCSC.utils_1.0.0 - [21] ggbeeswarm_0.7.2 purrr_1.0.2 - [23] bit_4.0.5 bluster_1.14.0 - [25] xfun_0.44 zlibbioc_1.50.0 - [27] cachem_1.1.0 beachmat_2.20.0 - [29] jsonlite_1.8.8 blob_1.2.4 - [31] highr_0.11 DelayedArray_0.30.1 - [33] BiocParallel_1.38.0 cluster_2.1.6 - [35] irlba_2.3.5.1 parallel_4.4.1 - [37] R6_2.5.1 Rcpp_1.0.12 - [39] knitr_1.47 splines_4.4.1 - [41] igraph_2.0.3 Matrix_1.7-0 - [43] tidyselect_1.2.1 viridis_0.6.5 - [45] abind_1.4-5 yaml_2.3.8 - [47] codetools_0.2-20 curl_5.2.1 - [49] lattice_0.22-6 tibble_3.2.1 - [51] withr_3.0.0 KEGGREST_1.44.0 - [53] BumpyMatrix_1.12.0 Rtsne_0.17 - [55] evaluate_0.23 BiocFileCache_2.12.0 - [57] ExperimentHub_2.12.0 Biostrings_2.72.1 - [59] pillar_1.9.0 BiocManager_1.30.23 - [61] filelock_1.0.3 renv_1.0.7 - [63] generics_0.1.3 BiocVersion_3.19.1 - [65] sparseMatrixStats_1.16.0 munsell_0.5.1 - [67] scales_1.3.0 glue_1.7.0 - [69] metapod_1.12.0 tools_4.4.1 - [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0 - [73] ScaledMatrix_1.12.0 locfit_1.5-9.9 - [75] cowplot_1.1.3 grid_4.4.1 - [77] AnnotationDbi_1.66.0 colorspace_2.1-0 - [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0 - [81] BiocSingular_1.20.0 vipor_0.4.7 - [83] cli_3.6.2 rsvd_1.0.5 - [85] rappdirs_0.3.3 fansi_1.0.6 - [87] viridisLite_0.4.2 S4Arrays_1.4.1 - [89] dplyr_1.1.4 ResidualMatrix_1.14.0 - [91] gtable_0.3.5 digest_0.6.35 - [93] dqrng_0.4.1 SparseArray_1.4.8 - [95] ggrepel_0.9.5 farver_2.1.2 - [97] rjson_0.2.21 memoise_2.0.1 - [99] htmltools_0.5.8.1 lifecycle_1.0.4 -[101] httr_1.4.7 statmod_1.5.0 -[103] mime_0.12 bit64_4.0.5 + [1] RColorBrewer_1.1-3 jsonlite_1.8.8 + [3] magrittr_2.0.3 ggbeeswarm_0.7.2 + [5] magick_2.8.3 farver_2.1.2 + [7] rmarkdown_2.27 zlibbioc_1.50.0 + [9] vctrs_0.6.5 memoise_2.0.1 + [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1 + [13] S4Arrays_1.4.1 AnnotationHub_3.12.0 + [15] curl_5.2.1 BiocNeighbors_1.22.0 + [17] SparseArray_1.4.8 cachem_1.1.0 + [19] ResidualMatrix_1.14.0 igraph_2.0.3 + [21] mime_0.12 lifecycle_1.0.4 + [23] pkgconfig_2.0.3 rsvd_1.0.5 + [25] Matrix_1.7-0 R6_2.5.1 + [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12 + [29] digest_0.6.35 colorspace_2.1-0 + [31] AnnotationDbi_1.66.0 dqrng_0.4.1 + [33] irlba_2.3.5.1 ExperimentHub_2.12.0 + [35] RSQLite_2.3.7 beachmat_2.20.0 + [37] filelock_1.0.3 labeling_0.4.3 + [39] fansi_1.0.6 httr_1.4.7 + [41] abind_1.4-5 compiler_4.4.1 + [43] bit64_4.0.5 withr_3.0.0 + [45] BiocParallel_1.38.0 viridis_0.6.5 + [47] DBI_1.2.3 highr_0.11 + [49] rappdirs_0.3.3 DelayedArray_0.30.1 + [51] rjson_0.2.21 bluster_1.14.0 + [53] tools_4.4.1 vipor_0.4.7 + [55] beeswarm_0.4.0 glue_1.7.0 + [57] grid_4.4.1 Rtsne_0.17 + [59] cluster_2.1.6 generics_0.1.3 + [61] gtable_0.3.5 BiocSingular_1.20.0 + [63] ScaledMatrix_1.12.0 metapod_1.12.0 + [65] utf8_1.2.4 XVector_0.44.0 + [67] ggrepel_0.9.5 BiocVersion_3.19.1 + [69] pillar_1.9.0 BumpyMatrix_1.12.0 + [71] splines_4.4.1 dplyr_1.1.4 + [73] BiocFileCache_2.12.0 lattice_0.22-6 + [75] renv_1.0.7 bit_4.0.5 + [77] tidyselect_1.2.1 locfit_1.5-9.9 + [79] Biostrings_2.72.1 knitr_1.47 + [81] gridExtra_2.3 xfun_0.44 + [83] statmod_1.5.0 UCSC.utils_1.0.0 + [85] yaml_2.3.8 evaluate_0.23 + [87] codetools_0.2-20 tibble_3.2.1 + [89] BiocManager_1.30.23 cli_3.6.2 + [91] munsell_0.5.1 Rcpp_1.0.12 + [93] dbplyr_2.5.0 png_0.1-8 + [95] parallel_4.4.1 blob_1.2.4 + [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2 + [99] scales_1.3.0 purrr_1.0.2 +[101] crayon_1.5.2 rlang_1.1.3 +[103] cowplot_1.1.3 KEGGREST_1.44.0 +[105] formatR_1.14
Exercises
@@ -4630,7 +4452,7 @@Exercise 1: Replicates
-Remember, you can subset SingleCellExperiments with logical indices, just like a matrix. You can also access their column data with the @@ -4644,7 +4466,7 @@
Give me a hint
-R
@@ -4652,11 +4474,12 @@ R
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
Exercise 2:
-You can just hand pheatmap() a matrix as its only
argument. It has a million options, but the defaults are usually pretty
@@ -4694,15 +4517,13 @@
Give me a hint
-R
-library(pheatmap)
-
-pheatmap(y.ab$counts)
+pheatmap(y.ab$counts)
The top DA result was a decrease in ExE ectoderm in the tomato condition, which you can sort of see, especially if you @@ -6462,7 +6283,7 @@
Exercise 1
-R
@@ -6495,7 +6316,7 @@ Exercise 2
-
+
R
@@ -6527,7 +6348,7 @@ Exercise 3
-
+
R
@@ -6568,7 +6389,7 @@ Exercise 4
-
+
R
diff --git a/cell_type_annotation.html b/cell_type_annotation.html
index 067a8e8..a423c5a 100644
--- a/cell_type_annotation.html
+++ b/cell_type_annotation.html
@@ -515,7 +515,7 @@ Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -666,7 +666,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -956,7 +956,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -1144,7 +1144,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -1153,231 +1155,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -1421,7 +1215,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -1566,7 +1360,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -1583,7 +1377,7 @@
Give me a hint
-
+
TODO
@@ -1612,7 +1406,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -1626,7 +1420,7 @@
Give me a hint
-
+
TODO
@@ -1656,7 +1450,7 @@ Exercise 3: Workflow
-
+
TODO
diff --git a/eda_qc.html b/eda_qc.html
index 6e7c4ba..f250bd7 100644
--- a/eda_qc.html
+++ b/eda_qc.html
@@ -486,7 +486,7 @@ Challenge
-
+
R
@@ -565,7 +565,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -791,7 +791,7 @@
Challenge
-
+
R
@@ -1063,7 +1063,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1349,7 +1349,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -1512,7 +1512,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
diff --git a/hca.html b/hca.html
index 18576b9..071b6cc 100644
--- a/hca.html
+++ b/hca.html
@@ -788,7 +788,7 @@
Exercise 1
-
+
R
@@ -821,7 +821,7 @@ Exercise 2
-
+
R
@@ -853,7 +853,7 @@ Exercise 3
-
+
R
@@ -894,7 +894,7 @@ Exercise 4
-
+
R
diff --git a/index.html b/index.html
index 24dd0f3..6dfcdaf 100644
--- a/index.html
+++ b/index.html
@@ -372,7 +372,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -398,7 +398,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN
website.
@@ -429,7 +429,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -453,7 +453,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN website.
- Select the
.pkg file for the latest R version
@@ -486,7 +486,7 @@
Install R using your package manager and RStudio
-
+
- Follow the instructions for your distribution from CRAN, they provide
information to get the most recent version of R for common
diff --git a/instructor/aio.html b/instructor/aio.html
index 4714d2d..f7a8835 100644
--- a/instructor/aio.html
+++ b/instructor/aio.html
@@ -1040,7 +1040,7 @@
Challenge
-
+
R
@@ -1120,7 +1120,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -1353,7 +1353,7 @@
Challenge
-
+
R
@@ -1629,7 +1629,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1924,7 +1924,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -2094,7 +2094,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
@@ -2380,7 +2380,7 @@
Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -2533,7 +2533,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -2826,7 +2826,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -3015,7 +3015,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -3024,231 +3026,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -3292,7 +3086,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -3441,7 +3235,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -3458,7 +3252,7 @@
Give me a hint
-
+
TODO
@@ -3487,7 +3281,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -3501,7 +3295,7 @@
Give me a hint
-
+
TODO
@@ -3531,7 +3325,7 @@ Exercise 3: Workflow
-
+
TODO
@@ -3603,7 +3397,7 @@ Key Points
-->
Content from Multi-sample analyses
-Last updated on 2024-09-06 |
+
Last updated on 2024-09-08 |
Edit this page
Estimated time: 45 minutes
@@ -3661,13 +3455,23 @@ Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.
We start by loading the data and doing a quick exploratory analysis,
essentially applying the normalization and visualization techniques that
-we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this
+time we’re selecting samples 5 to 10, not just 5 by itself.
R
library(MouseGastrulationData)
-sce <- WTChimeraData(samples=5:10, type = "processed")
+library(batchelor)
+library(edgeR)
+library(scater)
+library(ggplot2)
+library(scran)
+library(pheatmap)
+library(scuttle)
+
+sce <- WTChimeraData(samples = 5:10, type = "processed")
+
sce
@@ -3742,15 +3546,12 @@ OUTPUT<
R
-library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3758,23 +3559,24 @@ R
sce <- sce[,idx]
-We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps,
+and visualize them in a tSNE plot.
R
-# normalization
-sce <- logNormCounts(sce)
+sce <- logNormCounts(sce)
+
+dec <- modelGeneVar(sce, block = sce$sample)
-# identify highly variable genes
-dec <- modelGeneVar(sce, block=sce$sample)
chosen.hvgs <- dec$bio > 0
-# dimensionality reduction
sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000)
+
sce <- runTSNE(sce, dimred = "PCA")
sce$sample <- as.factor(sce$sample)
+
plotTSNE(sce, colour_by = "sample")
Challenge
-
+
Samples 5 and 6 were from the same “pool” of cells. Looking at the
documentation for the dataset under ?WTChimeraData we see
@@ -3822,7 +3624,7 @@
Show me the solution
first TSNE plot. If you were developing single-cell library preparation
protocols you might want to preserve this effect to understand how
variation in pools leads to variation in expression, but for now, given
-that we’re investigate other effects, we’ll want to remove this as
+that we’re investigating other effects, we’ll want to remove this as
undesired technical variation.
@@ -3831,29 +3633,30 @@ Show me the solution
Correcting batch effects
-We correct the effect of samples by aid of the
+
We “correct” the effect of samples with the
correctExperiment function in the batchelor
-package and using the sample colData column as
-batch.
+package and using the sample column as batch.
R
-library(batchelor)
-set.seed(10102)
-merged <- correctExperiments(sce,
- batch=sce$sample,
- subset.row=chosen.hvgs,
- PARAM=FastMnnParam(
- merge.order=list(
+set.seed(10102)
+
+merged <- correctExperiments(
+ sce,
+ batch = sce$sample,
+ subset.row = chosen.hvgs,
+ PARAM = FastMnnParam(
+ merge.order = list(
list(1,3,5), # WT (3 replicates)
list(2,4,6) # td-Tomato (3 replicates)
)
)
)
-merged <- runTSNE(merged, dimred="corrected")
-plotTSNE(merged, colour_by="batch")
+merged <- runTSNE(merged, dimred = "corrected")
+
+plotTSNE(merged, colour_by = "batch")
Once we removed the sample batch effect, we can proceed with the
Differential Expression Analysis.
@@ -3875,7 +3678,7 @@ Challenge
-
+
It’s important to have multiple samples within each experimental
group because it helps the batch effect correction algorithm distinguish
@@ -3898,7 +3701,7 @@
Show me the solution
Differential Expression
-In order to perform a Differential Expression Analysis, we need to
+
In order to perform a differential expression analysis, we need to
identify groups of cells across samples/conditions (depending on the
experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell
@@ -3926,9 +3729,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output
# corresponds to one unique combination of these two factors.
-library(scuttle)
-summed <- aggregateAcrossCells(merged,
- id=colData(merged)[,c("celltype.mapped", "sample")])
+
+summed <- aggregateAcrossCells(
+ merged,
+ id = colData(merged)[,c("celltype.mapped", "sample")]
+)
+
summed
R
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
y
@@ -4018,7 +3824,9 @@ R
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)
@@ -4034,8 +3842,10 @@ OUTPUT<
R
-keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)
@@ -4055,6 +3865,7 @@ R
y <- calcNormFactors(y)
+
y$samples
@@ -4095,7 +3906,7 @@ Rpar(mfrow = c(2,3))
for (i in seq_len(ncol(y))) {
- plotMD(y, column=i)
+ plotMD(y, column = i)
}
RR
-limma::plotMDS(cpm(y, log=TRUE),
- col=ifelse(y$samples$tomato, "red", "blue"))
+limma::plotMDS(cpm(y, log = TRUE),
+ col = ifelse(y$samples$tomato, "red", "blue"))
We then construct a design matrix by including both the pool and the
tomato as factors. This design indicates which samples belong to which
@@ -4153,6 +3964,7 @@
R
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)
@@ -4172,7 +3984,7 @@ RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized
-linear model for each gene. The robust=TRUE parameter
+linear model for each gene. The robust = TRUE parameter
avoids distortions from highly variable clusters. The QL method includes
an additional dispersion parameter, useful to handle the uncertainty and
variability of the per-gene variance, which is not well estimated by the
@@ -4182,7 +3994,8 @@
RR
-fit <- glmQLFit(y, design, robust=TRUE)
+fit <- glmQLFit(y, design, robust = TRUE)
+
summary(fit$var.prior)
@@ -4221,7 +4034,8 @@ RR
-res <- glmQLFTest(fit, coef=ncol(design))
+res <- glmQLFTest(fit, coef = ncol(design))
+
summary(decideTests(res))
@@ -4261,14 +4075,14 @@ OUTPUT<
R
-library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)
The returned object is a list of DataFrames each with
@@ -4280,6 +4094,7 @@
R
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]
Challenge
-
+
“logFC” stands for log fold-change. Rather than reporting e.g. a
5-fold increase, it’s better to report a logFC of log(5) = 1.61.
@@ -4336,26 +4151,29 @@
Show me the solution
With DA we look for differences in cluster abundance across
conditions (the tomato injection in our case), rather than differences
in gene expression.
-We first setup some code and variables for further analysis, like
-quantifying the number of cells per each cell type and fit a model to
-catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this
-time we start our analysis on the computed abundances and without
+
Our first steps are quantifying the number of cells per each cell
+type and fitting a model to catch differences between the injected cells
+and the background.
+The process is very similar differential expression modeling, but
+this time we start our analysis on the computed abundances and without
normalizing the data with TMM.
R
-library(edgeR)
-abundances <- table(merged$celltype.mapped, merged$sample)
+abundances <- table(merged$celltype.mapped, merged$sample)
+
abundances <- unclass(abundances)
-# Attaching some column metadata.
+
extra.info <- colData(merged)[match(colnames(abundances), merged$sample),]
-y.ab <- DGEList(abundances, samples=extra.info)
+
+y.ab <- DGEList(abundances, samples = extra.info)
design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples)
-y.ab <- estimateDisp(y.ab, design, trend="none")
-fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)
+
+y.ab <- estimateDisp(y.ab, design, trend = "none")
+
+fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)
Background on compositional effect
@@ -4401,6 +4219,7 @@ R
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
@@ -4413,11 +4232,11 @@ OUTPUT<
R
-y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))
@@ -4433,7 +4252,7 @@ OUTPUT<
R
-topTags(res2, n=10)
+topTags(res2, n = 10)
OUTPUT
@@ -4469,7 +4288,8 @@ Testing against a log-fold
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
@@ -4550,70 +4370,72 @@ OUTPUT<
[8] base
other attached packages:
- [1] edgeR_4.2.0 limma_3.60.2
- [3] batchelor_1.20.0 scran_1.32.0
- [5] scater_1.32.0 ggplot2_3.5.1
- [7] scuttle_1.14.0 MouseGastrulationData_1.18.0
- [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0
-[11] SummarizedExperiment_1.34.0 Biobase_2.64.0
-[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1
-[15] IRanges_2.38.0 S4Vectors_0.42.0
-[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0
-[19] matrixStats_1.3.0 BiocStyle_2.32.0
+ [1] pheatmap_1.0.12 scran_1.32.0
+ [3] scater_1.32.0 ggplot2_3.5.1
+ [5] scuttle_1.14.0 edgeR_4.2.0
+ [7] limma_3.60.2 batchelor_1.20.0
+ [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0
+[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0
+[13] Biobase_2.64.0 GenomicRanges_1.56.0
+[15] GenomeInfoDb_1.40.1 IRanges_2.38.0
+[17] S4Vectors_0.42.0 BiocGenerics_0.50.0
+[19] MatrixGenerics_1.16.0 matrixStats_1.3.0
+[21] BiocStyle_2.32.0
loaded via a namespace (and not attached):
- [1] DBI_1.2.3 formatR_1.14
- [3] gridExtra_2.3 rlang_1.1.3
- [5] magrittr_2.0.3 compiler_4.4.1
- [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0
- [9] png_0.1-8 vctrs_0.6.5
- [11] pkgconfig_2.0.3 crayon_1.5.2
- [13] fastmap_1.2.0 dbplyr_2.5.0
- [15] magick_2.8.3 XVector_0.44.0
- [17] labeling_0.4.3 utf8_1.2.4
- [19] rmarkdown_2.27 UCSC.utils_1.0.0
- [21] ggbeeswarm_0.7.2 purrr_1.0.2
- [23] bit_4.0.5 bluster_1.14.0
- [25] xfun_0.44 zlibbioc_1.50.0
- [27] cachem_1.1.0 beachmat_2.20.0
- [29] jsonlite_1.8.8 blob_1.2.4
- [31] highr_0.11 DelayedArray_0.30.1
- [33] BiocParallel_1.38.0 cluster_2.1.6
- [35] irlba_2.3.5.1 parallel_4.4.1
- [37] R6_2.5.1 Rcpp_1.0.12
- [39] knitr_1.47 splines_4.4.1
- [41] igraph_2.0.3 Matrix_1.7-0
- [43] tidyselect_1.2.1 viridis_0.6.5
- [45] abind_1.4-5 yaml_2.3.8
- [47] codetools_0.2-20 curl_5.2.1
- [49] lattice_0.22-6 tibble_3.2.1
- [51] withr_3.0.0 KEGGREST_1.44.0
- [53] BumpyMatrix_1.12.0 Rtsne_0.17
- [55] evaluate_0.23 BiocFileCache_2.12.0
- [57] ExperimentHub_2.12.0 Biostrings_2.72.1
- [59] pillar_1.9.0 BiocManager_1.30.23
- [61] filelock_1.0.3 renv_1.0.7
- [63] generics_0.1.3 BiocVersion_3.19.1
- [65] sparseMatrixStats_1.16.0 munsell_0.5.1
- [67] scales_1.3.0 glue_1.7.0
- [69] metapod_1.12.0 tools_4.4.1
- [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0
- [73] ScaledMatrix_1.12.0 locfit_1.5-9.9
- [75] cowplot_1.1.3 grid_4.4.1
- [77] AnnotationDbi_1.66.0 colorspace_2.1-0
- [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0
- [81] BiocSingular_1.20.0 vipor_0.4.7
- [83] cli_3.6.2 rsvd_1.0.5
- [85] rappdirs_0.3.3 fansi_1.0.6
- [87] viridisLite_0.4.2 S4Arrays_1.4.1
- [89] dplyr_1.1.4 ResidualMatrix_1.14.0
- [91] gtable_0.3.5 digest_0.6.35
- [93] dqrng_0.4.1 SparseArray_1.4.8
- [95] ggrepel_0.9.5 farver_2.1.2
- [97] rjson_0.2.21 memoise_2.0.1
- [99] htmltools_0.5.8.1 lifecycle_1.0.4
-[101] httr_1.4.7 statmod_1.5.0
-[103] mime_0.12 bit64_4.0.5
+ [1] RColorBrewer_1.1-3 jsonlite_1.8.8
+ [3] magrittr_2.0.3 ggbeeswarm_0.7.2
+ [5] magick_2.8.3 farver_2.1.2
+ [7] rmarkdown_2.27 zlibbioc_1.50.0
+ [9] vctrs_0.6.5 memoise_2.0.1
+ [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1
+ [13] S4Arrays_1.4.1 AnnotationHub_3.12.0
+ [15] curl_5.2.1 BiocNeighbors_1.22.0
+ [17] SparseArray_1.4.8 cachem_1.1.0
+ [19] ResidualMatrix_1.14.0 igraph_2.0.3
+ [21] mime_0.12 lifecycle_1.0.4
+ [23] pkgconfig_2.0.3 rsvd_1.0.5
+ [25] Matrix_1.7-0 R6_2.5.1
+ [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12
+ [29] digest_0.6.35 colorspace_2.1-0
+ [31] AnnotationDbi_1.66.0 dqrng_0.4.1
+ [33] irlba_2.3.5.1 ExperimentHub_2.12.0
+ [35] RSQLite_2.3.7 beachmat_2.20.0
+ [37] filelock_1.0.3 labeling_0.4.3
+ [39] fansi_1.0.6 httr_1.4.7
+ [41] abind_1.4-5 compiler_4.4.1
+ [43] bit64_4.0.5 withr_3.0.0
+ [45] BiocParallel_1.38.0 viridis_0.6.5
+ [47] DBI_1.2.3 highr_0.11
+ [49] rappdirs_0.3.3 DelayedArray_0.30.1
+ [51] rjson_0.2.21 bluster_1.14.0
+ [53] tools_4.4.1 vipor_0.4.7
+ [55] beeswarm_0.4.0 glue_1.7.0
+ [57] grid_4.4.1 Rtsne_0.17
+ [59] cluster_2.1.6 generics_0.1.3
+ [61] gtable_0.3.5 BiocSingular_1.20.0
+ [63] ScaledMatrix_1.12.0 metapod_1.12.0
+ [65] utf8_1.2.4 XVector_0.44.0
+ [67] ggrepel_0.9.5 BiocVersion_3.19.1
+ [69] pillar_1.9.0 BumpyMatrix_1.12.0
+ [71] splines_4.4.1 dplyr_1.1.4
+ [73] BiocFileCache_2.12.0 lattice_0.22-6
+ [75] renv_1.0.7 bit_4.0.5
+ [77] tidyselect_1.2.1 locfit_1.5-9.9
+ [79] Biostrings_2.72.1 knitr_1.47
+ [81] gridExtra_2.3 xfun_0.44
+ [83] statmod_1.5.0 UCSC.utils_1.0.0
+ [85] yaml_2.3.8 evaluate_0.23
+ [87] codetools_0.2-20 tibble_3.2.1
+ [89] BiocManager_1.30.23 cli_3.6.2
+ [91] munsell_0.5.1 Rcpp_1.0.12
+ [93] dbplyr_2.5.0 png_0.1-8
+ [95] parallel_4.4.1 blob_1.2.4
+ [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2
+ [99] scales_1.3.0 purrr_1.0.2
+[101] crayon_1.5.2 rlang_1.1.3
+[103] cowplot_1.1.3 KEGGREST_1.44.0
+[105] formatR_1.14
Exercises
@@ -4636,7 +4458,7 @@ Exercise 1: Replicates
-
+
Remember, you can subset SingleCellExperiments with logical indices,
just like a matrix. You can also access their column data with the
@@ -4650,7 +4472,7 @@
Give me a hint
-
+
R
@@ -4658,11 +4480,12 @@ R
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
@@ -4686,7 +4509,7 @@ Exercise 2:
-
+
You can just hand pheatmap() a matrix as its only
argument. It has a million options, but the defaults are usually pretty
@@ -4700,15 +4523,13 @@
Give me a hint
-
+
R
-library(pheatmap)
-
-pheatmap(y.ab$counts)
+pheatmap(y.ab$counts)
The top DA result was a decrease in ExE ectoderm in the tomato
condition, which you can sort of see, especially if you
@@ -6470,7 +6291,7 @@
Exercise 1
-
+
R
@@ -6503,7 +6324,7 @@ Exercise 2
-
+
R
@@ -6535,7 +6356,7 @@ Exercise 3
-
+
R
@@ -6527,7 +6348,7 @@ Exercise 3
-
+
R
@@ -6568,7 +6389,7 @@ Exercise 4
-
+
R
diff --git a/cell_type_annotation.html b/cell_type_annotation.html
index 067a8e8..a423c5a 100644
--- a/cell_type_annotation.html
+++ b/cell_type_annotation.html
@@ -515,7 +515,7 @@ Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -666,7 +666,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -956,7 +956,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -1144,7 +1144,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -1153,231 +1155,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -1421,7 +1215,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -1566,7 +1360,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -1583,7 +1377,7 @@
Give me a hint
-
+
TODO
@@ -1612,7 +1406,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -1626,7 +1420,7 @@
Give me a hint
-
+
TODO
@@ -1656,7 +1450,7 @@ Exercise 3: Workflow
-
+
TODO
diff --git a/eda_qc.html b/eda_qc.html
index 6e7c4ba..f250bd7 100644
--- a/eda_qc.html
+++ b/eda_qc.html
@@ -486,7 +486,7 @@ Challenge
-
+
R
@@ -565,7 +565,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -791,7 +791,7 @@
Challenge
-
+
R
@@ -1063,7 +1063,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1349,7 +1349,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -1512,7 +1512,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
diff --git a/hca.html b/hca.html
index 18576b9..071b6cc 100644
--- a/hca.html
+++ b/hca.html
@@ -788,7 +788,7 @@
Exercise 1
-
+
R
@@ -821,7 +821,7 @@ Exercise 2
-
+
R
@@ -853,7 +853,7 @@ Exercise 3
-
+
R
@@ -894,7 +894,7 @@ Exercise 4
-
+
R
diff --git a/index.html b/index.html
index 24dd0f3..6dfcdaf 100644
--- a/index.html
+++ b/index.html
@@ -372,7 +372,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -398,7 +398,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN
website.
@@ -429,7 +429,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -453,7 +453,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN website.
- Select the
.pkg file for the latest R version
@@ -486,7 +486,7 @@
Install R using your package manager and RStudio
-
+
- Follow the instructions for your distribution from CRAN, they provide
information to get the most recent version of R for common
diff --git a/instructor/aio.html b/instructor/aio.html
index 4714d2d..f7a8835 100644
--- a/instructor/aio.html
+++ b/instructor/aio.html
@@ -1040,7 +1040,7 @@
Challenge
-
+
R
@@ -1120,7 +1120,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -1353,7 +1353,7 @@
Challenge
-
+
R
@@ -1629,7 +1629,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1924,7 +1924,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -2094,7 +2094,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
@@ -2380,7 +2380,7 @@
Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -2533,7 +2533,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -2826,7 +2826,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -3015,7 +3015,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -3024,231 +3026,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -3292,7 +3086,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -3441,7 +3235,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -3458,7 +3252,7 @@
Give me a hint
-
+
TODO
@@ -3487,7 +3281,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -3501,7 +3295,7 @@
Give me a hint
-
+
TODO
@@ -3531,7 +3325,7 @@ Exercise 3: Workflow
-
+
TODO
@@ -3603,7 +3397,7 @@ Key Points
-->
Content from Multi-sample analyses
-Last updated on 2024-09-06 |
+
Last updated on 2024-09-08 |
Edit this page
Estimated time: 45 minutes
@@ -3661,13 +3455,23 @@ Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.
We start by loading the data and doing a quick exploratory analysis,
essentially applying the normalization and visualization techniques that
-we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this
+time we’re selecting samples 5 to 10, not just 5 by itself.
R
library(MouseGastrulationData)
-sce <- WTChimeraData(samples=5:10, type = "processed")
+library(batchelor)
+library(edgeR)
+library(scater)
+library(ggplot2)
+library(scran)
+library(pheatmap)
+library(scuttle)
+
+sce <- WTChimeraData(samples = 5:10, type = "processed")
+
sce
@@ -3742,15 +3546,12 @@ OUTPUT<
R
-library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3758,23 +3559,24 @@ R
sce <- sce[,idx]
-We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps,
+and visualize them in a tSNE plot.
R
-# normalization
-sce <- logNormCounts(sce)
+sce <- logNormCounts(sce)
+
+dec <- modelGeneVar(sce, block = sce$sample)
-# identify highly variable genes
-dec <- modelGeneVar(sce, block=sce$sample)
chosen.hvgs <- dec$bio > 0
-# dimensionality reduction
sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000)
+
sce <- runTSNE(sce, dimred = "PCA")
sce$sample <- as.factor(sce$sample)
+
plotTSNE(sce, colour_by = "sample")
Challenge
-
+
Samples 5 and 6 were from the same “pool” of cells. Looking at the
documentation for the dataset under ?WTChimeraData we see
@@ -3822,7 +3624,7 @@
Show me the solution
first TSNE plot. If you were developing single-cell library preparation
protocols you might want to preserve this effect to understand how
variation in pools leads to variation in expression, but for now, given
-that we’re investigate other effects, we’ll want to remove this as
+that we’re investigating other effects, we’ll want to remove this as
undesired technical variation.
@@ -3831,29 +3633,30 @@ Show me the solution
Correcting batch effects
-We correct the effect of samples by aid of the
+
We “correct” the effect of samples with the
correctExperiment function in the batchelor
-package and using the sample colData column as
-batch.
+package and using the sample column as batch.
R
-library(batchelor)
-set.seed(10102)
-merged <- correctExperiments(sce,
- batch=sce$sample,
- subset.row=chosen.hvgs,
- PARAM=FastMnnParam(
- merge.order=list(
+set.seed(10102)
+
+merged <- correctExperiments(
+ sce,
+ batch = sce$sample,
+ subset.row = chosen.hvgs,
+ PARAM = FastMnnParam(
+ merge.order = list(
list(1,3,5), # WT (3 replicates)
list(2,4,6) # td-Tomato (3 replicates)
)
)
)
-merged <- runTSNE(merged, dimred="corrected")
-plotTSNE(merged, colour_by="batch")
+merged <- runTSNE(merged, dimred = "corrected")
+
+plotTSNE(merged, colour_by = "batch")
Once we removed the sample batch effect, we can proceed with the
Differential Expression Analysis.
@@ -3875,7 +3678,7 @@ Challenge
-
+
It’s important to have multiple samples within each experimental
group because it helps the batch effect correction algorithm distinguish
@@ -3898,7 +3701,7 @@
Show me the solution
Differential Expression
-In order to perform a Differential Expression Analysis, we need to
+
In order to perform a differential expression analysis, we need to
identify groups of cells across samples/conditions (depending on the
experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell
@@ -3926,9 +3729,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output
# corresponds to one unique combination of these two factors.
-library(scuttle)
-summed <- aggregateAcrossCells(merged,
- id=colData(merged)[,c("celltype.mapped", "sample")])
+
+summed <- aggregateAcrossCells(
+ merged,
+ id = colData(merged)[,c("celltype.mapped", "sample")]
+)
+
summed
R
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
y
@@ -4018,7 +3824,9 @@ R
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)
@@ -4034,8 +3842,10 @@ OUTPUT<
R
-keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)
@@ -4055,6 +3865,7 @@ R
y <- calcNormFactors(y)
+
y$samples
@@ -4095,7 +3906,7 @@ Rpar(mfrow = c(2,3))
for (i in seq_len(ncol(y))) {
- plotMD(y, column=i)
+ plotMD(y, column = i)
}
RR
-limma::plotMDS(cpm(y, log=TRUE),
- col=ifelse(y$samples$tomato, "red", "blue"))
+limma::plotMDS(cpm(y, log = TRUE),
+ col = ifelse(y$samples$tomato, "red", "blue"))
We then construct a design matrix by including both the pool and the
tomato as factors. This design indicates which samples belong to which
@@ -4153,6 +3964,7 @@
R
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)
@@ -4172,7 +3984,7 @@ RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized
-linear model for each gene. The robust=TRUE parameter
+linear model for each gene. The robust = TRUE parameter
avoids distortions from highly variable clusters. The QL method includes
an additional dispersion parameter, useful to handle the uncertainty and
variability of the per-gene variance, which is not well estimated by the
@@ -4182,7 +3994,8 @@
RR
-fit <- glmQLFit(y, design, robust=TRUE)
+fit <- glmQLFit(y, design, robust = TRUE)
+
summary(fit$var.prior)
@@ -4221,7 +4034,8 @@ RR
-res <- glmQLFTest(fit, coef=ncol(design))
+res <- glmQLFTest(fit, coef = ncol(design))
+
summary(decideTests(res))
@@ -4261,14 +4075,14 @@ OUTPUT<
R
-library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)
The returned object is a list of DataFrames each with
@@ -4280,6 +4094,7 @@
R
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]
Challenge
-
+
“logFC” stands for log fold-change. Rather than reporting e.g. a
5-fold increase, it’s better to report a logFC of log(5) = 1.61.
@@ -4336,26 +4151,29 @@
Show me the solution
With DA we look for differences in cluster abundance across
conditions (the tomato injection in our case), rather than differences
in gene expression.
-We first setup some code and variables for further analysis, like
-quantifying the number of cells per each cell type and fit a model to
-catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this
-time we start our analysis on the computed abundances and without
+
Our first steps are quantifying the number of cells per each cell
+type and fitting a model to catch differences between the injected cells
+and the background.
+The process is very similar differential expression modeling, but
+this time we start our analysis on the computed abundances and without
normalizing the data with TMM.
R
-library(edgeR)
-abundances <- table(merged$celltype.mapped, merged$sample)
+abundances <- table(merged$celltype.mapped, merged$sample)
+
abundances <- unclass(abundances)
-# Attaching some column metadata.
+
extra.info <- colData(merged)[match(colnames(abundances), merged$sample),]
-y.ab <- DGEList(abundances, samples=extra.info)
+
+y.ab <- DGEList(abundances, samples = extra.info)
design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples)
-y.ab <- estimateDisp(y.ab, design, trend="none")
-fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)
+
+y.ab <- estimateDisp(y.ab, design, trend = "none")
+
+fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)
Background on compositional effect
@@ -4401,6 +4219,7 @@ R
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
@@ -4413,11 +4232,11 @@ OUTPUT<
R
-y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))
@@ -4433,7 +4252,7 @@ OUTPUT<
R
-topTags(res2, n=10)
+topTags(res2, n = 10)
OUTPUT
@@ -4469,7 +4288,8 @@ Testing against a log-fold
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
@@ -4550,70 +4370,72 @@ OUTPUT<
[8] base
other attached packages:
- [1] edgeR_4.2.0 limma_3.60.2
- [3] batchelor_1.20.0 scran_1.32.0
- [5] scater_1.32.0 ggplot2_3.5.1
- [7] scuttle_1.14.0 MouseGastrulationData_1.18.0
- [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0
-[11] SummarizedExperiment_1.34.0 Biobase_2.64.0
-[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1
-[15] IRanges_2.38.0 S4Vectors_0.42.0
-[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0
-[19] matrixStats_1.3.0 BiocStyle_2.32.0
+ [1] pheatmap_1.0.12 scran_1.32.0
+ [3] scater_1.32.0 ggplot2_3.5.1
+ [5] scuttle_1.14.0 edgeR_4.2.0
+ [7] limma_3.60.2 batchelor_1.20.0
+ [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0
+[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0
+[13] Biobase_2.64.0 GenomicRanges_1.56.0
+[15] GenomeInfoDb_1.40.1 IRanges_2.38.0
+[17] S4Vectors_0.42.0 BiocGenerics_0.50.0
+[19] MatrixGenerics_1.16.0 matrixStats_1.3.0
+[21] BiocStyle_2.32.0
loaded via a namespace (and not attached):
- [1] DBI_1.2.3 formatR_1.14
- [3] gridExtra_2.3 rlang_1.1.3
- [5] magrittr_2.0.3 compiler_4.4.1
- [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0
- [9] png_0.1-8 vctrs_0.6.5
- [11] pkgconfig_2.0.3 crayon_1.5.2
- [13] fastmap_1.2.0 dbplyr_2.5.0
- [15] magick_2.8.3 XVector_0.44.0
- [17] labeling_0.4.3 utf8_1.2.4
- [19] rmarkdown_2.27 UCSC.utils_1.0.0
- [21] ggbeeswarm_0.7.2 purrr_1.0.2
- [23] bit_4.0.5 bluster_1.14.0
- [25] xfun_0.44 zlibbioc_1.50.0
- [27] cachem_1.1.0 beachmat_2.20.0
- [29] jsonlite_1.8.8 blob_1.2.4
- [31] highr_0.11 DelayedArray_0.30.1
- [33] BiocParallel_1.38.0 cluster_2.1.6
- [35] irlba_2.3.5.1 parallel_4.4.1
- [37] R6_2.5.1 Rcpp_1.0.12
- [39] knitr_1.47 splines_4.4.1
- [41] igraph_2.0.3 Matrix_1.7-0
- [43] tidyselect_1.2.1 viridis_0.6.5
- [45] abind_1.4-5 yaml_2.3.8
- [47] codetools_0.2-20 curl_5.2.1
- [49] lattice_0.22-6 tibble_3.2.1
- [51] withr_3.0.0 KEGGREST_1.44.0
- [53] BumpyMatrix_1.12.0 Rtsne_0.17
- [55] evaluate_0.23 BiocFileCache_2.12.0
- [57] ExperimentHub_2.12.0 Biostrings_2.72.1
- [59] pillar_1.9.0 BiocManager_1.30.23
- [61] filelock_1.0.3 renv_1.0.7
- [63] generics_0.1.3 BiocVersion_3.19.1
- [65] sparseMatrixStats_1.16.0 munsell_0.5.1
- [67] scales_1.3.0 glue_1.7.0
- [69] metapod_1.12.0 tools_4.4.1
- [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0
- [73] ScaledMatrix_1.12.0 locfit_1.5-9.9
- [75] cowplot_1.1.3 grid_4.4.1
- [77] AnnotationDbi_1.66.0 colorspace_2.1-0
- [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0
- [81] BiocSingular_1.20.0 vipor_0.4.7
- [83] cli_3.6.2 rsvd_1.0.5
- [85] rappdirs_0.3.3 fansi_1.0.6
- [87] viridisLite_0.4.2 S4Arrays_1.4.1
- [89] dplyr_1.1.4 ResidualMatrix_1.14.0
- [91] gtable_0.3.5 digest_0.6.35
- [93] dqrng_0.4.1 SparseArray_1.4.8
- [95] ggrepel_0.9.5 farver_2.1.2
- [97] rjson_0.2.21 memoise_2.0.1
- [99] htmltools_0.5.8.1 lifecycle_1.0.4
-[101] httr_1.4.7 statmod_1.5.0
-[103] mime_0.12 bit64_4.0.5
+ [1] RColorBrewer_1.1-3 jsonlite_1.8.8
+ [3] magrittr_2.0.3 ggbeeswarm_0.7.2
+ [5] magick_2.8.3 farver_2.1.2
+ [7] rmarkdown_2.27 zlibbioc_1.50.0
+ [9] vctrs_0.6.5 memoise_2.0.1
+ [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1
+ [13] S4Arrays_1.4.1 AnnotationHub_3.12.0
+ [15] curl_5.2.1 BiocNeighbors_1.22.0
+ [17] SparseArray_1.4.8 cachem_1.1.0
+ [19] ResidualMatrix_1.14.0 igraph_2.0.3
+ [21] mime_0.12 lifecycle_1.0.4
+ [23] pkgconfig_2.0.3 rsvd_1.0.5
+ [25] Matrix_1.7-0 R6_2.5.1
+ [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12
+ [29] digest_0.6.35 colorspace_2.1-0
+ [31] AnnotationDbi_1.66.0 dqrng_0.4.1
+ [33] irlba_2.3.5.1 ExperimentHub_2.12.0
+ [35] RSQLite_2.3.7 beachmat_2.20.0
+ [37] filelock_1.0.3 labeling_0.4.3
+ [39] fansi_1.0.6 httr_1.4.7
+ [41] abind_1.4-5 compiler_4.4.1
+ [43] bit64_4.0.5 withr_3.0.0
+ [45] BiocParallel_1.38.0 viridis_0.6.5
+ [47] DBI_1.2.3 highr_0.11
+ [49] rappdirs_0.3.3 DelayedArray_0.30.1
+ [51] rjson_0.2.21 bluster_1.14.0
+ [53] tools_4.4.1 vipor_0.4.7
+ [55] beeswarm_0.4.0 glue_1.7.0
+ [57] grid_4.4.1 Rtsne_0.17
+ [59] cluster_2.1.6 generics_0.1.3
+ [61] gtable_0.3.5 BiocSingular_1.20.0
+ [63] ScaledMatrix_1.12.0 metapod_1.12.0
+ [65] utf8_1.2.4 XVector_0.44.0
+ [67] ggrepel_0.9.5 BiocVersion_3.19.1
+ [69] pillar_1.9.0 BumpyMatrix_1.12.0
+ [71] splines_4.4.1 dplyr_1.1.4
+ [73] BiocFileCache_2.12.0 lattice_0.22-6
+ [75] renv_1.0.7 bit_4.0.5
+ [77] tidyselect_1.2.1 locfit_1.5-9.9
+ [79] Biostrings_2.72.1 knitr_1.47
+ [81] gridExtra_2.3 xfun_0.44
+ [83] statmod_1.5.0 UCSC.utils_1.0.0
+ [85] yaml_2.3.8 evaluate_0.23
+ [87] codetools_0.2-20 tibble_3.2.1
+ [89] BiocManager_1.30.23 cli_3.6.2
+ [91] munsell_0.5.1 Rcpp_1.0.12
+ [93] dbplyr_2.5.0 png_0.1-8
+ [95] parallel_4.4.1 blob_1.2.4
+ [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2
+ [99] scales_1.3.0 purrr_1.0.2
+[101] crayon_1.5.2 rlang_1.1.3
+[103] cowplot_1.1.3 KEGGREST_1.44.0
+[105] formatR_1.14
Exercises
@@ -4636,7 +4458,7 @@ Exercise 1: Replicates
-
+
Remember, you can subset SingleCellExperiments with logical indices,
just like a matrix. You can also access their column data with the
@@ -4650,7 +4472,7 @@
Give me a hint
-
+
R
@@ -4658,11 +4480,12 @@ R
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
@@ -4686,7 +4509,7 @@ Exercise 2:
-
+
You can just hand pheatmap() a matrix as its only
argument. It has a million options, but the defaults are usually pretty
@@ -4700,15 +4523,13 @@
Give me a hint
-
+
R
-library(pheatmap)
-
-pheatmap(y.ab$counts)
+pheatmap(y.ab$counts)
The top DA result was a decrease in ExE ectoderm in the tomato
condition, which you can sort of see, especially if you
@@ -6470,7 +6291,7 @@
Exercise 1
-
+
R
@@ -6503,7 +6324,7 @@ Exercise 2
-
+
R
@@ -6535,7 +6356,7 @@ Exercise 3
-
+
R
@@ -6568,7 +6389,7 @@ Exercise 4
-
+
R
diff --git a/cell_type_annotation.html b/cell_type_annotation.html
index 067a8e8..a423c5a 100644
--- a/cell_type_annotation.html
+++ b/cell_type_annotation.html
@@ -515,7 +515,7 @@ Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -666,7 +666,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -956,7 +956,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -1144,7 +1144,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -1153,231 +1155,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -1421,7 +1215,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -1566,7 +1360,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -1583,7 +1377,7 @@
Give me a hint
-
+
TODO
@@ -1612,7 +1406,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -1626,7 +1420,7 @@
Give me a hint
-
+
TODO
@@ -1656,7 +1450,7 @@ Exercise 3: Workflow
-
+
TODO
diff --git a/eda_qc.html b/eda_qc.html
index 6e7c4ba..f250bd7 100644
--- a/eda_qc.html
+++ b/eda_qc.html
@@ -486,7 +486,7 @@ Challenge
-
+
R
@@ -565,7 +565,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -791,7 +791,7 @@
Challenge
-
+
R
@@ -1063,7 +1063,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1349,7 +1349,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -1512,7 +1512,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
diff --git a/hca.html b/hca.html
index 18576b9..071b6cc 100644
--- a/hca.html
+++ b/hca.html
@@ -788,7 +788,7 @@
Exercise 1
-
+
R
@@ -821,7 +821,7 @@ Exercise 2
-
+
R
@@ -853,7 +853,7 @@ Exercise 3
-
+
R
@@ -894,7 +894,7 @@ Exercise 4
-
+
R
diff --git a/index.html b/index.html
index 24dd0f3..6dfcdaf 100644
--- a/index.html
+++ b/index.html
@@ -372,7 +372,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -398,7 +398,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN
website.
@@ -429,7 +429,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -453,7 +453,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN website.
- Select the
.pkg file for the latest R version
@@ -486,7 +486,7 @@
Install R using your package manager and RStudio
-
+
- Follow the instructions for your distribution from CRAN, they provide
information to get the most recent version of R for common
diff --git a/instructor/aio.html b/instructor/aio.html
index 4714d2d..f7a8835 100644
--- a/instructor/aio.html
+++ b/instructor/aio.html
@@ -1040,7 +1040,7 @@
Challenge
-
+
R
@@ -1120,7 +1120,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -1353,7 +1353,7 @@
Challenge
-
+
R
@@ -1629,7 +1629,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1924,7 +1924,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -2094,7 +2094,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
@@ -2380,7 +2380,7 @@
Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -2533,7 +2533,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -2826,7 +2826,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -3015,7 +3015,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -3024,231 +3026,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -3292,7 +3086,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -3441,7 +3235,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -3458,7 +3252,7 @@
Give me a hint
-
+
TODO
@@ -3487,7 +3281,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -3501,7 +3295,7 @@
Give me a hint
-
+
TODO
@@ -3531,7 +3325,7 @@ Exercise 3: Workflow
-
+
TODO
@@ -3603,7 +3397,7 @@ Key Points
-->
Content from Multi-sample analyses
-Last updated on 2024-09-06 |
+
Last updated on 2024-09-08 |
Edit this page
Estimated time: 45 minutes
@@ -3661,13 +3455,23 @@ Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.
We start by loading the data and doing a quick exploratory analysis,
essentially applying the normalization and visualization techniques that
-we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this
+time we’re selecting samples 5 to 10, not just 5 by itself.
R
library(MouseGastrulationData)
-sce <- WTChimeraData(samples=5:10, type = "processed")
+library(batchelor)
+library(edgeR)
+library(scater)
+library(ggplot2)
+library(scran)
+library(pheatmap)
+library(scuttle)
+
+sce <- WTChimeraData(samples = 5:10, type = "processed")
+
sce
@@ -3742,15 +3546,12 @@ OUTPUT<
R
-library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3758,23 +3559,24 @@ R
sce <- sce[,idx]
-We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps,
+and visualize them in a tSNE plot.
R
-# normalization
-sce <- logNormCounts(sce)
+sce <- logNormCounts(sce)
+
+dec <- modelGeneVar(sce, block = sce$sample)
-# identify highly variable genes
-dec <- modelGeneVar(sce, block=sce$sample)
chosen.hvgs <- dec$bio > 0
-# dimensionality reduction
sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000)
+
sce <- runTSNE(sce, dimred = "PCA")
sce$sample <- as.factor(sce$sample)
+
plotTSNE(sce, colour_by = "sample")
Challenge
-
+
Samples 5 and 6 were from the same “pool” of cells. Looking at the
documentation for the dataset under ?WTChimeraData we see
@@ -3822,7 +3624,7 @@
Show me the solution
first TSNE plot. If you were developing single-cell library preparation
protocols you might want to preserve this effect to understand how
variation in pools leads to variation in expression, but for now, given
-that we’re investigate other effects, we’ll want to remove this as
+that we’re investigating other effects, we’ll want to remove this as
undesired technical variation.
@@ -3831,29 +3633,30 @@ Show me the solution
Correcting batch effects
-We correct the effect of samples by aid of the
+
We “correct” the effect of samples with the
correctExperiment function in the batchelor
-package and using the sample colData column as
-batch.
+package and using the sample column as batch.
R
-library(batchelor)
-set.seed(10102)
-merged <- correctExperiments(sce,
- batch=sce$sample,
- subset.row=chosen.hvgs,
- PARAM=FastMnnParam(
- merge.order=list(
+set.seed(10102)
+
+merged <- correctExperiments(
+ sce,
+ batch = sce$sample,
+ subset.row = chosen.hvgs,
+ PARAM = FastMnnParam(
+ merge.order = list(
list(1,3,5), # WT (3 replicates)
list(2,4,6) # td-Tomato (3 replicates)
)
)
)
-merged <- runTSNE(merged, dimred="corrected")
-plotTSNE(merged, colour_by="batch")
+merged <- runTSNE(merged, dimred = "corrected")
+
+plotTSNE(merged, colour_by = "batch")
Once we removed the sample batch effect, we can proceed with the
Differential Expression Analysis.
@@ -3875,7 +3678,7 @@ Challenge
-
+
It’s important to have multiple samples within each experimental
group because it helps the batch effect correction algorithm distinguish
@@ -3898,7 +3701,7 @@
Show me the solution
Differential Expression
-In order to perform a Differential Expression Analysis, we need to
+
In order to perform a differential expression analysis, we need to
identify groups of cells across samples/conditions (depending on the
experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell
@@ -3926,9 +3729,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output
# corresponds to one unique combination of these two factors.
-library(scuttle)
-summed <- aggregateAcrossCells(merged,
- id=colData(merged)[,c("celltype.mapped", "sample")])
+
+summed <- aggregateAcrossCells(
+ merged,
+ id = colData(merged)[,c("celltype.mapped", "sample")]
+)
+
summed
R
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
y
@@ -4018,7 +3824,9 @@ R
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)
@@ -4034,8 +3842,10 @@ OUTPUT<
R
-keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)
@@ -4055,6 +3865,7 @@ R
y <- calcNormFactors(y)
+
y$samples
@@ -4095,7 +3906,7 @@ Rpar(mfrow = c(2,3))
for (i in seq_len(ncol(y))) {
- plotMD(y, column=i)
+ plotMD(y, column = i)
}
RR
-limma::plotMDS(cpm(y, log=TRUE),
- col=ifelse(y$samples$tomato, "red", "blue"))
+limma::plotMDS(cpm(y, log = TRUE),
+ col = ifelse(y$samples$tomato, "red", "blue"))
We then construct a design matrix by including both the pool and the
tomato as factors. This design indicates which samples belong to which
@@ -4153,6 +3964,7 @@
R
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)
@@ -4172,7 +3984,7 @@ RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized
-linear model for each gene. The robust=TRUE parameter
+linear model for each gene. The robust = TRUE parameter
avoids distortions from highly variable clusters. The QL method includes
an additional dispersion parameter, useful to handle the uncertainty and
variability of the per-gene variance, which is not well estimated by the
@@ -4182,7 +3994,8 @@
RR
-fit <- glmQLFit(y, design, robust=TRUE)
+fit <- glmQLFit(y, design, robust = TRUE)
+
summary(fit$var.prior)
@@ -4221,7 +4034,8 @@ RR
-res <- glmQLFTest(fit, coef=ncol(design))
+res <- glmQLFTest(fit, coef = ncol(design))
+
summary(decideTests(res))
@@ -4261,14 +4075,14 @@ OUTPUT<
R
-library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)
The returned object is a list of DataFrames each with
@@ -4280,6 +4094,7 @@
R
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]
Challenge
-
+
“logFC” stands for log fold-change. Rather than reporting e.g. a
5-fold increase, it’s better to report a logFC of log(5) = 1.61.
@@ -4336,26 +4151,29 @@
Show me the solution
With DA we look for differences in cluster abundance across
conditions (the tomato injection in our case), rather than differences
in gene expression.
-We first setup some code and variables for further analysis, like
-quantifying the number of cells per each cell type and fit a model to
-catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this
-time we start our analysis on the computed abundances and without
+
Our first steps are quantifying the number of cells per each cell
+type and fitting a model to catch differences between the injected cells
+and the background.
+The process is very similar differential expression modeling, but
+this time we start our analysis on the computed abundances and without
normalizing the data with TMM.
R
-library(edgeR)
-abundances <- table(merged$celltype.mapped, merged$sample)
+abundances <- table(merged$celltype.mapped, merged$sample)
+
abundances <- unclass(abundances)
-# Attaching some column metadata.
+
extra.info <- colData(merged)[match(colnames(abundances), merged$sample),]
-y.ab <- DGEList(abundances, samples=extra.info)
+
+y.ab <- DGEList(abundances, samples = extra.info)
design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples)
-y.ab <- estimateDisp(y.ab, design, trend="none")
-fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)
+
+y.ab <- estimateDisp(y.ab, design, trend = "none")
+
+fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)
Background on compositional effect
@@ -4401,6 +4219,7 @@ R
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
@@ -4413,11 +4232,11 @@ OUTPUT<
R
-y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))
@@ -4433,7 +4252,7 @@ OUTPUT<
R
-topTags(res2, n=10)
+topTags(res2, n = 10)
OUTPUT
@@ -4469,7 +4288,8 @@ Testing against a log-fold
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
@@ -4550,70 +4370,72 @@ OUTPUT<
[8] base
other attached packages:
- [1] edgeR_4.2.0 limma_3.60.2
- [3] batchelor_1.20.0 scran_1.32.0
- [5] scater_1.32.0 ggplot2_3.5.1
- [7] scuttle_1.14.0 MouseGastrulationData_1.18.0
- [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0
-[11] SummarizedExperiment_1.34.0 Biobase_2.64.0
-[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1
-[15] IRanges_2.38.0 S4Vectors_0.42.0
-[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0
-[19] matrixStats_1.3.0 BiocStyle_2.32.0
+ [1] pheatmap_1.0.12 scran_1.32.0
+ [3] scater_1.32.0 ggplot2_3.5.1
+ [5] scuttle_1.14.0 edgeR_4.2.0
+ [7] limma_3.60.2 batchelor_1.20.0
+ [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0
+[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0
+[13] Biobase_2.64.0 GenomicRanges_1.56.0
+[15] GenomeInfoDb_1.40.1 IRanges_2.38.0
+[17] S4Vectors_0.42.0 BiocGenerics_0.50.0
+[19] MatrixGenerics_1.16.0 matrixStats_1.3.0
+[21] BiocStyle_2.32.0
loaded via a namespace (and not attached):
- [1] DBI_1.2.3 formatR_1.14
- [3] gridExtra_2.3 rlang_1.1.3
- [5] magrittr_2.0.3 compiler_4.4.1
- [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0
- [9] png_0.1-8 vctrs_0.6.5
- [11] pkgconfig_2.0.3 crayon_1.5.2
- [13] fastmap_1.2.0 dbplyr_2.5.0
- [15] magick_2.8.3 XVector_0.44.0
- [17] labeling_0.4.3 utf8_1.2.4
- [19] rmarkdown_2.27 UCSC.utils_1.0.0
- [21] ggbeeswarm_0.7.2 purrr_1.0.2
- [23] bit_4.0.5 bluster_1.14.0
- [25] xfun_0.44 zlibbioc_1.50.0
- [27] cachem_1.1.0 beachmat_2.20.0
- [29] jsonlite_1.8.8 blob_1.2.4
- [31] highr_0.11 DelayedArray_0.30.1
- [33] BiocParallel_1.38.0 cluster_2.1.6
- [35] irlba_2.3.5.1 parallel_4.4.1
- [37] R6_2.5.1 Rcpp_1.0.12
- [39] knitr_1.47 splines_4.4.1
- [41] igraph_2.0.3 Matrix_1.7-0
- [43] tidyselect_1.2.1 viridis_0.6.5
- [45] abind_1.4-5 yaml_2.3.8
- [47] codetools_0.2-20 curl_5.2.1
- [49] lattice_0.22-6 tibble_3.2.1
- [51] withr_3.0.0 KEGGREST_1.44.0
- [53] BumpyMatrix_1.12.0 Rtsne_0.17
- [55] evaluate_0.23 BiocFileCache_2.12.0
- [57] ExperimentHub_2.12.0 Biostrings_2.72.1
- [59] pillar_1.9.0 BiocManager_1.30.23
- [61] filelock_1.0.3 renv_1.0.7
- [63] generics_0.1.3 BiocVersion_3.19.1
- [65] sparseMatrixStats_1.16.0 munsell_0.5.1
- [67] scales_1.3.0 glue_1.7.0
- [69] metapod_1.12.0 tools_4.4.1
- [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0
- [73] ScaledMatrix_1.12.0 locfit_1.5-9.9
- [75] cowplot_1.1.3 grid_4.4.1
- [77] AnnotationDbi_1.66.0 colorspace_2.1-0
- [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0
- [81] BiocSingular_1.20.0 vipor_0.4.7
- [83] cli_3.6.2 rsvd_1.0.5
- [85] rappdirs_0.3.3 fansi_1.0.6
- [87] viridisLite_0.4.2 S4Arrays_1.4.1
- [89] dplyr_1.1.4 ResidualMatrix_1.14.0
- [91] gtable_0.3.5 digest_0.6.35
- [93] dqrng_0.4.1 SparseArray_1.4.8
- [95] ggrepel_0.9.5 farver_2.1.2
- [97] rjson_0.2.21 memoise_2.0.1
- [99] htmltools_0.5.8.1 lifecycle_1.0.4
-[101] httr_1.4.7 statmod_1.5.0
-[103] mime_0.12 bit64_4.0.5
+ [1] RColorBrewer_1.1-3 jsonlite_1.8.8
+ [3] magrittr_2.0.3 ggbeeswarm_0.7.2
+ [5] magick_2.8.3 farver_2.1.2
+ [7] rmarkdown_2.27 zlibbioc_1.50.0
+ [9] vctrs_0.6.5 memoise_2.0.1
+ [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1
+ [13] S4Arrays_1.4.1 AnnotationHub_3.12.0
+ [15] curl_5.2.1 BiocNeighbors_1.22.0
+ [17] SparseArray_1.4.8 cachem_1.1.0
+ [19] ResidualMatrix_1.14.0 igraph_2.0.3
+ [21] mime_0.12 lifecycle_1.0.4
+ [23] pkgconfig_2.0.3 rsvd_1.0.5
+ [25] Matrix_1.7-0 R6_2.5.1
+ [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12
+ [29] digest_0.6.35 colorspace_2.1-0
+ [31] AnnotationDbi_1.66.0 dqrng_0.4.1
+ [33] irlba_2.3.5.1 ExperimentHub_2.12.0
+ [35] RSQLite_2.3.7 beachmat_2.20.0
+ [37] filelock_1.0.3 labeling_0.4.3
+ [39] fansi_1.0.6 httr_1.4.7
+ [41] abind_1.4-5 compiler_4.4.1
+ [43] bit64_4.0.5 withr_3.0.0
+ [45] BiocParallel_1.38.0 viridis_0.6.5
+ [47] DBI_1.2.3 highr_0.11
+ [49] rappdirs_0.3.3 DelayedArray_0.30.1
+ [51] rjson_0.2.21 bluster_1.14.0
+ [53] tools_4.4.1 vipor_0.4.7
+ [55] beeswarm_0.4.0 glue_1.7.0
+ [57] grid_4.4.1 Rtsne_0.17
+ [59] cluster_2.1.6 generics_0.1.3
+ [61] gtable_0.3.5 BiocSingular_1.20.0
+ [63] ScaledMatrix_1.12.0 metapod_1.12.0
+ [65] utf8_1.2.4 XVector_0.44.0
+ [67] ggrepel_0.9.5 BiocVersion_3.19.1
+ [69] pillar_1.9.0 BumpyMatrix_1.12.0
+ [71] splines_4.4.1 dplyr_1.1.4
+ [73] BiocFileCache_2.12.0 lattice_0.22-6
+ [75] renv_1.0.7 bit_4.0.5
+ [77] tidyselect_1.2.1 locfit_1.5-9.9
+ [79] Biostrings_2.72.1 knitr_1.47
+ [81] gridExtra_2.3 xfun_0.44
+ [83] statmod_1.5.0 UCSC.utils_1.0.0
+ [85] yaml_2.3.8 evaluate_0.23
+ [87] codetools_0.2-20 tibble_3.2.1
+ [89] BiocManager_1.30.23 cli_3.6.2
+ [91] munsell_0.5.1 Rcpp_1.0.12
+ [93] dbplyr_2.5.0 png_0.1-8
+ [95] parallel_4.4.1 blob_1.2.4
+ [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2
+ [99] scales_1.3.0 purrr_1.0.2
+[101] crayon_1.5.2 rlang_1.1.3
+[103] cowplot_1.1.3 KEGGREST_1.44.0
+[105] formatR_1.14
Exercises
@@ -4636,7 +4458,7 @@ Exercise 1: Replicates
-
+
Remember, you can subset SingleCellExperiments with logical indices,
just like a matrix. You can also access their column data with the
@@ -4650,7 +4472,7 @@
Give me a hint
-
+
R
@@ -4658,11 +4480,12 @@ R
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
@@ -4686,7 +4509,7 @@ Exercise 2:
-
+
You can just hand pheatmap() a matrix as its only
argument. It has a million options, but the defaults are usually pretty
@@ -4700,15 +4523,13 @@
Give me a hint
-
+
R
-library(pheatmap)
-
-pheatmap(y.ab$counts)
+pheatmap(y.ab$counts)
The top DA result was a decrease in ExE ectoderm in the tomato
condition, which you can sort of see, especially if you
@@ -6470,7 +6291,7 @@
Exercise 1
-
+
R
@@ -6503,7 +6324,7 @@ Exercise 2
-
+
R
@@ -6535,7 +6356,7 @@ Exercise 3
-
+
R
diff --git a/cell_type_annotation.html b/cell_type_annotation.html
index 067a8e8..a423c5a 100644
--- a/cell_type_annotation.html
+++ b/cell_type_annotation.html
@@ -515,7 +515,7 @@ Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -666,7 +666,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -956,7 +956,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -1144,7 +1144,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -1153,231 +1155,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -1421,7 +1215,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -1566,7 +1360,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -1583,7 +1377,7 @@
Give me a hint
-
+
TODO
@@ -1612,7 +1406,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -1626,7 +1420,7 @@
Give me a hint
-
+
TODO
@@ -1656,7 +1450,7 @@ Exercise 3: Workflow
-
+
TODO
diff --git a/eda_qc.html b/eda_qc.html
index 6e7c4ba..f250bd7 100644
--- a/eda_qc.html
+++ b/eda_qc.html
@@ -486,7 +486,7 @@ Challenge
-
+
R
@@ -565,7 +565,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -791,7 +791,7 @@
Challenge
-
+
R
@@ -1063,7 +1063,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1349,7 +1349,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -1512,7 +1512,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
diff --git a/hca.html b/hca.html
index 18576b9..071b6cc 100644
--- a/hca.html
+++ b/hca.html
@@ -788,7 +788,7 @@
Exercise 1
-
+
R
@@ -821,7 +821,7 @@ Exercise 2
-
+
R
@@ -853,7 +853,7 @@ Exercise 3
-
+
R
@@ -894,7 +894,7 @@ Exercise 4
-
+
R
diff --git a/index.html b/index.html
index 24dd0f3..6dfcdaf 100644
--- a/index.html
+++ b/index.html
@@ -372,7 +372,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -398,7 +398,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN
website.
@@ -429,7 +429,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -453,7 +453,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN website.
- Select the
.pkg file for the latest R version
@@ -486,7 +486,7 @@
Install R using your package manager and RStudio
-
+
- Follow the instructions for your distribution from CRAN, they provide
information to get the most recent version of R for common
diff --git a/instructor/aio.html b/instructor/aio.html
index 4714d2d..f7a8835 100644
--- a/instructor/aio.html
+++ b/instructor/aio.html
@@ -1040,7 +1040,7 @@
Challenge
-
+
R
@@ -1120,7 +1120,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -1353,7 +1353,7 @@
Challenge
-
+
R
@@ -1629,7 +1629,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1924,7 +1924,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -2094,7 +2094,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
@@ -2380,7 +2380,7 @@
Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -2533,7 +2533,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -2826,7 +2826,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -3015,7 +3015,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -3024,231 +3026,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -3292,7 +3086,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -3441,7 +3235,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -3458,7 +3252,7 @@
Give me a hint
-
+
TODO
@@ -3487,7 +3281,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -3501,7 +3295,7 @@
Give me a hint
-
+
TODO
@@ -3531,7 +3325,7 @@ Exercise 3: Workflow
-
+
TODO
@@ -3603,7 +3397,7 @@ Key Points
-->
Content from Multi-sample analyses
-Last updated on 2024-09-06 |
+
Last updated on 2024-09-08 |
Edit this page
Estimated time: 45 minutes
@@ -3661,13 +3455,23 @@ Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.
We start by loading the data and doing a quick exploratory analysis,
essentially applying the normalization and visualization techniques that
-we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this
+time we’re selecting samples 5 to 10, not just 5 by itself.
R
library(MouseGastrulationData)
-sce <- WTChimeraData(samples=5:10, type = "processed")
+library(batchelor)
+library(edgeR)
+library(scater)
+library(ggplot2)
+library(scran)
+library(pheatmap)
+library(scuttle)
+
+sce <- WTChimeraData(samples = 5:10, type = "processed")
+
sce
@@ -3742,15 +3546,12 @@ OUTPUT<
R
-library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3758,23 +3559,24 @@ R
sce <- sce[,idx]
-We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps,
+and visualize them in a tSNE plot.
R
-# normalization
-sce <- logNormCounts(sce)
+sce <- logNormCounts(sce)
+
+dec <- modelGeneVar(sce, block = sce$sample)
-# identify highly variable genes
-dec <- modelGeneVar(sce, block=sce$sample)
chosen.hvgs <- dec$bio > 0
-# dimensionality reduction
sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000)
+
sce <- runTSNE(sce, dimred = "PCA")
sce$sample <- as.factor(sce$sample)
+
plotTSNE(sce, colour_by = "sample")
Challenge
-
+
Samples 5 and 6 were from the same “pool” of cells. Looking at the
documentation for the dataset under ?WTChimeraData we see
@@ -3822,7 +3624,7 @@
Show me the solution
first TSNE plot. If you were developing single-cell library preparation
protocols you might want to preserve this effect to understand how
variation in pools leads to variation in expression, but for now, given
-that we’re investigate other effects, we’ll want to remove this as
+that we’re investigating other effects, we’ll want to remove this as
undesired technical variation.
@@ -3831,29 +3633,30 @@ Show me the solution
Correcting batch effects
-We correct the effect of samples by aid of the
+
We “correct” the effect of samples with the
correctExperiment function in the batchelor
-package and using the sample colData column as
-batch.
+package and using the sample column as batch.
R
-library(batchelor)
-set.seed(10102)
-merged <- correctExperiments(sce,
- batch=sce$sample,
- subset.row=chosen.hvgs,
- PARAM=FastMnnParam(
- merge.order=list(
+set.seed(10102)
+
+merged <- correctExperiments(
+ sce,
+ batch = sce$sample,
+ subset.row = chosen.hvgs,
+ PARAM = FastMnnParam(
+ merge.order = list(
list(1,3,5), # WT (3 replicates)
list(2,4,6) # td-Tomato (3 replicates)
)
)
)
-merged <- runTSNE(merged, dimred="corrected")
-plotTSNE(merged, colour_by="batch")
+merged <- runTSNE(merged, dimred = "corrected")
+
+plotTSNE(merged, colour_by = "batch")
Once we removed the sample batch effect, we can proceed with the
Differential Expression Analysis.
@@ -3875,7 +3678,7 @@ Challenge
-
+
It’s important to have multiple samples within each experimental
group because it helps the batch effect correction algorithm distinguish
@@ -3898,7 +3701,7 @@
Show me the solution
Differential Expression
-In order to perform a Differential Expression Analysis, we need to
+
In order to perform a differential expression analysis, we need to
identify groups of cells across samples/conditions (depending on the
experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell
@@ -3926,9 +3729,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output
# corresponds to one unique combination of these two factors.
-library(scuttle)
-summed <- aggregateAcrossCells(merged,
- id=colData(merged)[,c("celltype.mapped", "sample")])
+
+summed <- aggregateAcrossCells(
+ merged,
+ id = colData(merged)[,c("celltype.mapped", "sample")]
+)
+
summed
R
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
y
@@ -4018,7 +3824,9 @@ R
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)
@@ -4034,8 +3842,10 @@ OUTPUT<
R
-keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)
@@ -4055,6 +3865,7 @@ R
y <- calcNormFactors(y)
+
y$samples
@@ -4095,7 +3906,7 @@ Rpar(mfrow = c(2,3))
for (i in seq_len(ncol(y))) {
- plotMD(y, column=i)
+ plotMD(y, column = i)
}
RR
-limma::plotMDS(cpm(y, log=TRUE),
- col=ifelse(y$samples$tomato, "red", "blue"))
+limma::plotMDS(cpm(y, log = TRUE),
+ col = ifelse(y$samples$tomato, "red", "blue"))
We then construct a design matrix by including both the pool and the
tomato as factors. This design indicates which samples belong to which
@@ -4153,6 +3964,7 @@
R
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)
@@ -4172,7 +3984,7 @@ RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized
-linear model for each gene. The robust=TRUE parameter
+linear model for each gene. The robust = TRUE parameter
avoids distortions from highly variable clusters. The QL method includes
an additional dispersion parameter, useful to handle the uncertainty and
variability of the per-gene variance, which is not well estimated by the
@@ -4182,7 +3994,8 @@
RR
-fit <- glmQLFit(y, design, robust=TRUE)
+fit <- glmQLFit(y, design, robust = TRUE)
+
summary(fit$var.prior)
@@ -4221,7 +4034,8 @@ RR
-res <- glmQLFTest(fit, coef=ncol(design))
+res <- glmQLFTest(fit, coef = ncol(design))
+
summary(decideTests(res))
@@ -4261,14 +4075,14 @@ OUTPUT<
R
-library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)
The returned object is a list of DataFrames each with
@@ -4280,6 +4094,7 @@
R
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]
Challenge
-
+
“logFC” stands for log fold-change. Rather than reporting e.g. a
5-fold increase, it’s better to report a logFC of log(5) = 1.61.
@@ -4336,26 +4151,29 @@
Show me the solution
With DA we look for differences in cluster abundance across
conditions (the tomato injection in our case), rather than differences
in gene expression.
-We first setup some code and variables for further analysis, like
-quantifying the number of cells per each cell type and fit a model to
-catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this
-time we start our analysis on the computed abundances and without
+
Our first steps are quantifying the number of cells per each cell
+type and fitting a model to catch differences between the injected cells
+and the background.
+The process is very similar differential expression modeling, but
+this time we start our analysis on the computed abundances and without
normalizing the data with TMM.
R
-library(edgeR)
-abundances <- table(merged$celltype.mapped, merged$sample)
+abundances <- table(merged$celltype.mapped, merged$sample)
+
abundances <- unclass(abundances)
-# Attaching some column metadata.
+
extra.info <- colData(merged)[match(colnames(abundances), merged$sample),]
-y.ab <- DGEList(abundances, samples=extra.info)
+
+y.ab <- DGEList(abundances, samples = extra.info)
design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples)
-y.ab <- estimateDisp(y.ab, design, trend="none")
-fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)
+
+y.ab <- estimateDisp(y.ab, design, trend = "none")
+
+fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)
Background on compositional effect
@@ -4401,6 +4219,7 @@ R
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
@@ -4413,11 +4232,11 @@ OUTPUT<
R
-y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))
@@ -4433,7 +4252,7 @@ OUTPUT<
R
-topTags(res2, n=10)
+topTags(res2, n = 10)
OUTPUT
@@ -4469,7 +4288,8 @@ Testing against a log-fold
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
@@ -4550,70 +4370,72 @@ OUTPUT<
[8] base
other attached packages:
- [1] edgeR_4.2.0 limma_3.60.2
- [3] batchelor_1.20.0 scran_1.32.0
- [5] scater_1.32.0 ggplot2_3.5.1
- [7] scuttle_1.14.0 MouseGastrulationData_1.18.0
- [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0
-[11] SummarizedExperiment_1.34.0 Biobase_2.64.0
-[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1
-[15] IRanges_2.38.0 S4Vectors_0.42.0
-[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0
-[19] matrixStats_1.3.0 BiocStyle_2.32.0
+ [1] pheatmap_1.0.12 scran_1.32.0
+ [3] scater_1.32.0 ggplot2_3.5.1
+ [5] scuttle_1.14.0 edgeR_4.2.0
+ [7] limma_3.60.2 batchelor_1.20.0
+ [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0
+[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0
+[13] Biobase_2.64.0 GenomicRanges_1.56.0
+[15] GenomeInfoDb_1.40.1 IRanges_2.38.0
+[17] S4Vectors_0.42.0 BiocGenerics_0.50.0
+[19] MatrixGenerics_1.16.0 matrixStats_1.3.0
+[21] BiocStyle_2.32.0
loaded via a namespace (and not attached):
- [1] DBI_1.2.3 formatR_1.14
- [3] gridExtra_2.3 rlang_1.1.3
- [5] magrittr_2.0.3 compiler_4.4.1
- [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0
- [9] png_0.1-8 vctrs_0.6.5
- [11] pkgconfig_2.0.3 crayon_1.5.2
- [13] fastmap_1.2.0 dbplyr_2.5.0
- [15] magick_2.8.3 XVector_0.44.0
- [17] labeling_0.4.3 utf8_1.2.4
- [19] rmarkdown_2.27 UCSC.utils_1.0.0
- [21] ggbeeswarm_0.7.2 purrr_1.0.2
- [23] bit_4.0.5 bluster_1.14.0
- [25] xfun_0.44 zlibbioc_1.50.0
- [27] cachem_1.1.0 beachmat_2.20.0
- [29] jsonlite_1.8.8 blob_1.2.4
- [31] highr_0.11 DelayedArray_0.30.1
- [33] BiocParallel_1.38.0 cluster_2.1.6
- [35] irlba_2.3.5.1 parallel_4.4.1
- [37] R6_2.5.1 Rcpp_1.0.12
- [39] knitr_1.47 splines_4.4.1
- [41] igraph_2.0.3 Matrix_1.7-0
- [43] tidyselect_1.2.1 viridis_0.6.5
- [45] abind_1.4-5 yaml_2.3.8
- [47] codetools_0.2-20 curl_5.2.1
- [49] lattice_0.22-6 tibble_3.2.1
- [51] withr_3.0.0 KEGGREST_1.44.0
- [53] BumpyMatrix_1.12.0 Rtsne_0.17
- [55] evaluate_0.23 BiocFileCache_2.12.0
- [57] ExperimentHub_2.12.0 Biostrings_2.72.1
- [59] pillar_1.9.0 BiocManager_1.30.23
- [61] filelock_1.0.3 renv_1.0.7
- [63] generics_0.1.3 BiocVersion_3.19.1
- [65] sparseMatrixStats_1.16.0 munsell_0.5.1
- [67] scales_1.3.0 glue_1.7.0
- [69] metapod_1.12.0 tools_4.4.1
- [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0
- [73] ScaledMatrix_1.12.0 locfit_1.5-9.9
- [75] cowplot_1.1.3 grid_4.4.1
- [77] AnnotationDbi_1.66.0 colorspace_2.1-0
- [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0
- [81] BiocSingular_1.20.0 vipor_0.4.7
- [83] cli_3.6.2 rsvd_1.0.5
- [85] rappdirs_0.3.3 fansi_1.0.6
- [87] viridisLite_0.4.2 S4Arrays_1.4.1
- [89] dplyr_1.1.4 ResidualMatrix_1.14.0
- [91] gtable_0.3.5 digest_0.6.35
- [93] dqrng_0.4.1 SparseArray_1.4.8
- [95] ggrepel_0.9.5 farver_2.1.2
- [97] rjson_0.2.21 memoise_2.0.1
- [99] htmltools_0.5.8.1 lifecycle_1.0.4
-[101] httr_1.4.7 statmod_1.5.0
-[103] mime_0.12 bit64_4.0.5
+ [1] RColorBrewer_1.1-3 jsonlite_1.8.8
+ [3] magrittr_2.0.3 ggbeeswarm_0.7.2
+ [5] magick_2.8.3 farver_2.1.2
+ [7] rmarkdown_2.27 zlibbioc_1.50.0
+ [9] vctrs_0.6.5 memoise_2.0.1
+ [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1
+ [13] S4Arrays_1.4.1 AnnotationHub_3.12.0
+ [15] curl_5.2.1 BiocNeighbors_1.22.0
+ [17] SparseArray_1.4.8 cachem_1.1.0
+ [19] ResidualMatrix_1.14.0 igraph_2.0.3
+ [21] mime_0.12 lifecycle_1.0.4
+ [23] pkgconfig_2.0.3 rsvd_1.0.5
+ [25] Matrix_1.7-0 R6_2.5.1
+ [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12
+ [29] digest_0.6.35 colorspace_2.1-0
+ [31] AnnotationDbi_1.66.0 dqrng_0.4.1
+ [33] irlba_2.3.5.1 ExperimentHub_2.12.0
+ [35] RSQLite_2.3.7 beachmat_2.20.0
+ [37] filelock_1.0.3 labeling_0.4.3
+ [39] fansi_1.0.6 httr_1.4.7
+ [41] abind_1.4-5 compiler_4.4.1
+ [43] bit64_4.0.5 withr_3.0.0
+ [45] BiocParallel_1.38.0 viridis_0.6.5
+ [47] DBI_1.2.3 highr_0.11
+ [49] rappdirs_0.3.3 DelayedArray_0.30.1
+ [51] rjson_0.2.21 bluster_1.14.0
+ [53] tools_4.4.1 vipor_0.4.7
+ [55] beeswarm_0.4.0 glue_1.7.0
+ [57] grid_4.4.1 Rtsne_0.17
+ [59] cluster_2.1.6 generics_0.1.3
+ [61] gtable_0.3.5 BiocSingular_1.20.0
+ [63] ScaledMatrix_1.12.0 metapod_1.12.0
+ [65] utf8_1.2.4 XVector_0.44.0
+ [67] ggrepel_0.9.5 BiocVersion_3.19.1
+ [69] pillar_1.9.0 BumpyMatrix_1.12.0
+ [71] splines_4.4.1 dplyr_1.1.4
+ [73] BiocFileCache_2.12.0 lattice_0.22-6
+ [75] renv_1.0.7 bit_4.0.5
+ [77] tidyselect_1.2.1 locfit_1.5-9.9
+ [79] Biostrings_2.72.1 knitr_1.47
+ [81] gridExtra_2.3 xfun_0.44
+ [83] statmod_1.5.0 UCSC.utils_1.0.0
+ [85] yaml_2.3.8 evaluate_0.23
+ [87] codetools_0.2-20 tibble_3.2.1
+ [89] BiocManager_1.30.23 cli_3.6.2
+ [91] munsell_0.5.1 Rcpp_1.0.12
+ [93] dbplyr_2.5.0 png_0.1-8
+ [95] parallel_4.4.1 blob_1.2.4
+ [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2
+ [99] scales_1.3.0 purrr_1.0.2
+[101] crayon_1.5.2 rlang_1.1.3
+[103] cowplot_1.1.3 KEGGREST_1.44.0
+[105] formatR_1.14
Exercises
@@ -4636,7 +4458,7 @@ Exercise 1: Replicates
-
+
Remember, you can subset SingleCellExperiments with logical indices,
just like a matrix. You can also access their column data with the
@@ -4650,7 +4472,7 @@
Give me a hint
-
+
R
@@ -4658,11 +4480,12 @@ R
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
@@ -4686,7 +4509,7 @@ Exercise 2:
-
+
You can just hand pheatmap() a matrix as its only
argument. It has a million options, but the defaults are usually pretty
@@ -4700,15 +4523,13 @@
Give me a hint
-
+
R
-library(pheatmap)
-
-pheatmap(y.ab$counts)
+pheatmap(y.ab$counts)
The top DA result was a decrease in ExE ectoderm in the tomato
condition, which you can sort of see, especially if you
@@ -6470,7 +6291,7 @@
Exercise 1
-
+
R
@@ -6503,7 +6324,7 @@ Exercise 2
-
+
R
@@ -6535,7 +6356,7 @@ Exercise 3
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -666,7 +666,7 @@
Challenge
-One important reason why is because averages over all other clusters can be sensitive to the cell type composition. If a rare cell type shows @@ -956,7 +956,7 @@
Challenge
-Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -1144,7 +1144,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -1153,231 +1155,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
R
@@ -1153,231 +1155,23 @@R tab <- table(new.labels, sce$celltype.mapped) -tab +tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20As a diagnostic measure, we examine the distribution of AUCs across cells for each label. In heterogeneous populations, the distribution for @@ -1421,7 +1215,7 @@
Challenge
-The example that jumps out most strongly to the eye is ExE endoderm, which doesn’t show clear separate modes. Simultaneously, Endothelium @@ -1566,7 +1360,7 @@
Exercise 1: Clustering
-The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -1583,7 +1377,7 @@
Give me a hint
-TODO
Exercise 2: Cluster annotation
-Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -1626,7 +1420,7 @@
Give me a hint
-TODO
Exercise 3: Workflow
-TODO
Challenge
-R
@@ -565,7 +565,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -791,7 +791,7 @@
Challenge
-
+
R
@@ -1063,7 +1063,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1349,7 +1349,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -1512,7 +1512,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
diff --git a/hca.html b/hca.html
index 18576b9..071b6cc 100644
--- a/hca.html
+++ b/hca.html
@@ -788,7 +788,7 @@
Exercise 1
-
+
R
@@ -821,7 +821,7 @@ Exercise 2
-
+
R
@@ -853,7 +853,7 @@ Exercise 3
-
+
R
@@ -894,7 +894,7 @@ Exercise 4
-
+
R
diff --git a/index.html b/index.html
index 24dd0f3..6dfcdaf 100644
--- a/index.html
+++ b/index.html
@@ -372,7 +372,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -398,7 +398,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN
website.
@@ -429,7 +429,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -453,7 +453,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN website.
- Select the
.pkg file for the latest R version
@@ -486,7 +486,7 @@
Install R using your package manager and RStudio
-
+
- Follow the instructions for your distribution from CRAN, they provide
information to get the most recent version of R for common
diff --git a/instructor/aio.html b/instructor/aio.html
index 4714d2d..f7a8835 100644
--- a/instructor/aio.html
+++ b/instructor/aio.html
@@ -1040,7 +1040,7 @@
Challenge
-
+
R
@@ -1120,7 +1120,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -1353,7 +1353,7 @@
Challenge
-
+
R
@@ -1629,7 +1629,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1924,7 +1924,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -2094,7 +2094,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
@@ -2380,7 +2380,7 @@
Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -2533,7 +2533,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -2826,7 +2826,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -3015,7 +3015,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -3024,231 +3026,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -3292,7 +3086,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -3441,7 +3235,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -3458,7 +3252,7 @@
Give me a hint
-
+
TODO
@@ -3487,7 +3281,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -3501,7 +3295,7 @@
Give me a hint
-
+
TODO
@@ -3531,7 +3325,7 @@ Exercise 3: Workflow
-
+
TODO
@@ -3603,7 +3397,7 @@ Key Points
-->
Content from Multi-sample analyses
-Last updated on 2024-09-06 |
+
Last updated on 2024-09-08 |
Edit this page
Estimated time: 45 minutes
@@ -3661,13 +3455,23 @@ Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.
We start by loading the data and doing a quick exploratory analysis,
essentially applying the normalization and visualization techniques that
-we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this
+time we’re selecting samples 5 to 10, not just 5 by itself.
R
library(MouseGastrulationData)
-sce <- WTChimeraData(samples=5:10, type = "processed")
+library(batchelor)
+library(edgeR)
+library(scater)
+library(ggplot2)
+library(scran)
+library(pheatmap)
+library(scuttle)
+
+sce <- WTChimeraData(samples = 5:10, type = "processed")
+
sce
@@ -3742,15 +3546,12 @@ OUTPUT<
R
-library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3758,23 +3559,24 @@ R
sce <- sce[,idx]
-We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps,
+and visualize them in a tSNE plot.
R
-# normalization
-sce <- logNormCounts(sce)
+sce <- logNormCounts(sce)
+
+dec <- modelGeneVar(sce, block = sce$sample)
-# identify highly variable genes
-dec <- modelGeneVar(sce, block=sce$sample)
chosen.hvgs <- dec$bio > 0
-# dimensionality reduction
sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000)
+
sce <- runTSNE(sce, dimred = "PCA")
sce$sample <- as.factor(sce$sample)
+
plotTSNE(sce, colour_by = "sample")
Challenge
-
+
Samples 5 and 6 were from the same “pool” of cells. Looking at the
documentation for the dataset under ?WTChimeraData we see
@@ -3822,7 +3624,7 @@
Show me the solution
first TSNE plot. If you were developing single-cell library preparation
protocols you might want to preserve this effect to understand how
variation in pools leads to variation in expression, but for now, given
-that we’re investigate other effects, we’ll want to remove this as
+that we’re investigating other effects, we’ll want to remove this as
undesired technical variation.
@@ -3831,29 +3633,30 @@ Show me the solution
Correcting batch effects
-We correct the effect of samples by aid of the
+
We “correct” the effect of samples with the
correctExperiment function in the batchelor
-package and using the sample colData column as
-batch.
+package and using the sample column as batch.
R
-library(batchelor)
-set.seed(10102)
-merged <- correctExperiments(sce,
- batch=sce$sample,
- subset.row=chosen.hvgs,
- PARAM=FastMnnParam(
- merge.order=list(
+set.seed(10102)
+
+merged <- correctExperiments(
+ sce,
+ batch = sce$sample,
+ subset.row = chosen.hvgs,
+ PARAM = FastMnnParam(
+ merge.order = list(
list(1,3,5), # WT (3 replicates)
list(2,4,6) # td-Tomato (3 replicates)
)
)
)
-merged <- runTSNE(merged, dimred="corrected")
-plotTSNE(merged, colour_by="batch")
+merged <- runTSNE(merged, dimred = "corrected")
+
+plotTSNE(merged, colour_by = "batch")
Once we removed the sample batch effect, we can proceed with the
Differential Expression Analysis.
@@ -3875,7 +3678,7 @@ Challenge
-
+
It’s important to have multiple samples within each experimental
group because it helps the batch effect correction algorithm distinguish
@@ -3898,7 +3701,7 @@
Show me the solution
Differential Expression
-In order to perform a Differential Expression Analysis, we need to
+
In order to perform a differential expression analysis, we need to
identify groups of cells across samples/conditions (depending on the
experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell
@@ -3926,9 +3729,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output
# corresponds to one unique combination of these two factors.
-library(scuttle)
-summed <- aggregateAcrossCells(merged,
- id=colData(merged)[,c("celltype.mapped", "sample")])
+
+summed <- aggregateAcrossCells(
+ merged,
+ id = colData(merged)[,c("celltype.mapped", "sample")]
+)
+
summed
R
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
y
@@ -4018,7 +3824,9 @@ R
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)
@@ -4034,8 +3842,10 @@ OUTPUT<
R
-keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)
@@ -4055,6 +3865,7 @@ R
y <- calcNormFactors(y)
+
y$samples
@@ -4095,7 +3906,7 @@ Rpar(mfrow = c(2,3))
for (i in seq_len(ncol(y))) {
- plotMD(y, column=i)
+ plotMD(y, column = i)
}
RR
-limma::plotMDS(cpm(y, log=TRUE),
- col=ifelse(y$samples$tomato, "red", "blue"))
+limma::plotMDS(cpm(y, log = TRUE),
+ col = ifelse(y$samples$tomato, "red", "blue"))
We then construct a design matrix by including both the pool and the
tomato as factors. This design indicates which samples belong to which
@@ -4153,6 +3964,7 @@
R
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)
@@ -4172,7 +3984,7 @@ RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized
-linear model for each gene. The robust=TRUE parameter
+linear model for each gene. The robust = TRUE parameter
avoids distortions from highly variable clusters. The QL method includes
an additional dispersion parameter, useful to handle the uncertainty and
variability of the per-gene variance, which is not well estimated by the
@@ -4182,7 +3994,8 @@
RR
-fit <- glmQLFit(y, design, robust=TRUE)
+fit <- glmQLFit(y, design, robust = TRUE)
+
summary(fit$var.prior)
@@ -4221,7 +4034,8 @@ RR
-res <- glmQLFTest(fit, coef=ncol(design))
+res <- glmQLFTest(fit, coef = ncol(design))
+
summary(decideTests(res))
@@ -4261,14 +4075,14 @@ OUTPUT<
R
-library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)
The returned object is a list of DataFrames each with
@@ -4280,6 +4094,7 @@
R
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]
Challenge
-
+
“logFC” stands for log fold-change. Rather than reporting e.g. a
5-fold increase, it’s better to report a logFC of log(5) = 1.61.
@@ -4336,26 +4151,29 @@
Show me the solution
With DA we look for differences in cluster abundance across
conditions (the tomato injection in our case), rather than differences
in gene expression.
-We first setup some code and variables for further analysis, like
-quantifying the number of cells per each cell type and fit a model to
-catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this
-time we start our analysis on the computed abundances and without
+
Our first steps are quantifying the number of cells per each cell
+type and fitting a model to catch differences between the injected cells
+and the background.
+The process is very similar differential expression modeling, but
+this time we start our analysis on the computed abundances and without
normalizing the data with TMM.
R
-library(edgeR)
-abundances <- table(merged$celltype.mapped, merged$sample)
+abundances <- table(merged$celltype.mapped, merged$sample)
+
abundances <- unclass(abundances)
-# Attaching some column metadata.
+
extra.info <- colData(merged)[match(colnames(abundances), merged$sample),]
-y.ab <- DGEList(abundances, samples=extra.info)
+
+y.ab <- DGEList(abundances, samples = extra.info)
design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples)
-y.ab <- estimateDisp(y.ab, design, trend="none")
-fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)
+
+y.ab <- estimateDisp(y.ab, design, trend = "none")
+
+fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)
Background on compositional effect
@@ -4401,6 +4219,7 @@ R
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
@@ -4413,11 +4232,11 @@ OUTPUT<
R
-y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))
@@ -4433,7 +4252,7 @@ OUTPUT<
R
-topTags(res2, n=10)
+topTags(res2, n = 10)
OUTPUT
@@ -4469,7 +4288,8 @@ Testing against a log-fold
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
@@ -4550,70 +4370,72 @@ OUTPUT<
[8] base
other attached packages:
- [1] edgeR_4.2.0 limma_3.60.2
- [3] batchelor_1.20.0 scran_1.32.0
- [5] scater_1.32.0 ggplot2_3.5.1
- [7] scuttle_1.14.0 MouseGastrulationData_1.18.0
- [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0
-[11] SummarizedExperiment_1.34.0 Biobase_2.64.0
-[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1
-[15] IRanges_2.38.0 S4Vectors_0.42.0
-[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0
-[19] matrixStats_1.3.0 BiocStyle_2.32.0
+ [1] pheatmap_1.0.12 scran_1.32.0
+ [3] scater_1.32.0 ggplot2_3.5.1
+ [5] scuttle_1.14.0 edgeR_4.2.0
+ [7] limma_3.60.2 batchelor_1.20.0
+ [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0
+[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0
+[13] Biobase_2.64.0 GenomicRanges_1.56.0
+[15] GenomeInfoDb_1.40.1 IRanges_2.38.0
+[17] S4Vectors_0.42.0 BiocGenerics_0.50.0
+[19] MatrixGenerics_1.16.0 matrixStats_1.3.0
+[21] BiocStyle_2.32.0
loaded via a namespace (and not attached):
- [1] DBI_1.2.3 formatR_1.14
- [3] gridExtra_2.3 rlang_1.1.3
- [5] magrittr_2.0.3 compiler_4.4.1
- [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0
- [9] png_0.1-8 vctrs_0.6.5
- [11] pkgconfig_2.0.3 crayon_1.5.2
- [13] fastmap_1.2.0 dbplyr_2.5.0
- [15] magick_2.8.3 XVector_0.44.0
- [17] labeling_0.4.3 utf8_1.2.4
- [19] rmarkdown_2.27 UCSC.utils_1.0.0
- [21] ggbeeswarm_0.7.2 purrr_1.0.2
- [23] bit_4.0.5 bluster_1.14.0
- [25] xfun_0.44 zlibbioc_1.50.0
- [27] cachem_1.1.0 beachmat_2.20.0
- [29] jsonlite_1.8.8 blob_1.2.4
- [31] highr_0.11 DelayedArray_0.30.1
- [33] BiocParallel_1.38.0 cluster_2.1.6
- [35] irlba_2.3.5.1 parallel_4.4.1
- [37] R6_2.5.1 Rcpp_1.0.12
- [39] knitr_1.47 splines_4.4.1
- [41] igraph_2.0.3 Matrix_1.7-0
- [43] tidyselect_1.2.1 viridis_0.6.5
- [45] abind_1.4-5 yaml_2.3.8
- [47] codetools_0.2-20 curl_5.2.1
- [49] lattice_0.22-6 tibble_3.2.1
- [51] withr_3.0.0 KEGGREST_1.44.0
- [53] BumpyMatrix_1.12.0 Rtsne_0.17
- [55] evaluate_0.23 BiocFileCache_2.12.0
- [57] ExperimentHub_2.12.0 Biostrings_2.72.1
- [59] pillar_1.9.0 BiocManager_1.30.23
- [61] filelock_1.0.3 renv_1.0.7
- [63] generics_0.1.3 BiocVersion_3.19.1
- [65] sparseMatrixStats_1.16.0 munsell_0.5.1
- [67] scales_1.3.0 glue_1.7.0
- [69] metapod_1.12.0 tools_4.4.1
- [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0
- [73] ScaledMatrix_1.12.0 locfit_1.5-9.9
- [75] cowplot_1.1.3 grid_4.4.1
- [77] AnnotationDbi_1.66.0 colorspace_2.1-0
- [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0
- [81] BiocSingular_1.20.0 vipor_0.4.7
- [83] cli_3.6.2 rsvd_1.0.5
- [85] rappdirs_0.3.3 fansi_1.0.6
- [87] viridisLite_0.4.2 S4Arrays_1.4.1
- [89] dplyr_1.1.4 ResidualMatrix_1.14.0
- [91] gtable_0.3.5 digest_0.6.35
- [93] dqrng_0.4.1 SparseArray_1.4.8
- [95] ggrepel_0.9.5 farver_2.1.2
- [97] rjson_0.2.21 memoise_2.0.1
- [99] htmltools_0.5.8.1 lifecycle_1.0.4
-[101] httr_1.4.7 statmod_1.5.0
-[103] mime_0.12 bit64_4.0.5
+ [1] RColorBrewer_1.1-3 jsonlite_1.8.8
+ [3] magrittr_2.0.3 ggbeeswarm_0.7.2
+ [5] magick_2.8.3 farver_2.1.2
+ [7] rmarkdown_2.27 zlibbioc_1.50.0
+ [9] vctrs_0.6.5 memoise_2.0.1
+ [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1
+ [13] S4Arrays_1.4.1 AnnotationHub_3.12.0
+ [15] curl_5.2.1 BiocNeighbors_1.22.0
+ [17] SparseArray_1.4.8 cachem_1.1.0
+ [19] ResidualMatrix_1.14.0 igraph_2.0.3
+ [21] mime_0.12 lifecycle_1.0.4
+ [23] pkgconfig_2.0.3 rsvd_1.0.5
+ [25] Matrix_1.7-0 R6_2.5.1
+ [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12
+ [29] digest_0.6.35 colorspace_2.1-0
+ [31] AnnotationDbi_1.66.0 dqrng_0.4.1
+ [33] irlba_2.3.5.1 ExperimentHub_2.12.0
+ [35] RSQLite_2.3.7 beachmat_2.20.0
+ [37] filelock_1.0.3 labeling_0.4.3
+ [39] fansi_1.0.6 httr_1.4.7
+ [41] abind_1.4-5 compiler_4.4.1
+ [43] bit64_4.0.5 withr_3.0.0
+ [45] BiocParallel_1.38.0 viridis_0.6.5
+ [47] DBI_1.2.3 highr_0.11
+ [49] rappdirs_0.3.3 DelayedArray_0.30.1
+ [51] rjson_0.2.21 bluster_1.14.0
+ [53] tools_4.4.1 vipor_0.4.7
+ [55] beeswarm_0.4.0 glue_1.7.0
+ [57] grid_4.4.1 Rtsne_0.17
+ [59] cluster_2.1.6 generics_0.1.3
+ [61] gtable_0.3.5 BiocSingular_1.20.0
+ [63] ScaledMatrix_1.12.0 metapod_1.12.0
+ [65] utf8_1.2.4 XVector_0.44.0
+ [67] ggrepel_0.9.5 BiocVersion_3.19.1
+ [69] pillar_1.9.0 BumpyMatrix_1.12.0
+ [71] splines_4.4.1 dplyr_1.1.4
+ [73] BiocFileCache_2.12.0 lattice_0.22-6
+ [75] renv_1.0.7 bit_4.0.5
+ [77] tidyselect_1.2.1 locfit_1.5-9.9
+ [79] Biostrings_2.72.1 knitr_1.47
+ [81] gridExtra_2.3 xfun_0.44
+ [83] statmod_1.5.0 UCSC.utils_1.0.0
+ [85] yaml_2.3.8 evaluate_0.23
+ [87] codetools_0.2-20 tibble_3.2.1
+ [89] BiocManager_1.30.23 cli_3.6.2
+ [91] munsell_0.5.1 Rcpp_1.0.12
+ [93] dbplyr_2.5.0 png_0.1-8
+ [95] parallel_4.4.1 blob_1.2.4
+ [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2
+ [99] scales_1.3.0 purrr_1.0.2
+[101] crayon_1.5.2 rlang_1.1.3
+[103] cowplot_1.1.3 KEGGREST_1.44.0
+[105] formatR_1.14
Exercises
@@ -4636,7 +4458,7 @@ Exercise 1: Replicates
-
+
Remember, you can subset SingleCellExperiments with logical indices,
just like a matrix. You can also access their column data with the
@@ -4650,7 +4472,7 @@
Give me a hint
-
+
R
@@ -4658,11 +4480,12 @@ R
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
@@ -4686,7 +4509,7 @@ Exercise 2:
-
+
You can just hand pheatmap() a matrix as its only
argument. It has a million options, but the defaults are usually pretty
@@ -4700,15 +4523,13 @@
Give me a hint
-
+
R
-library(pheatmap)
-
-pheatmap(y.ab$counts)
+pheatmap(y.ab$counts)
The top DA result was a decrease in ExE ectoderm in the tomato
condition, which you can sort of see, especially if you
@@ -6470,7 +6291,7 @@
Exercise 1
-
+
R
@@ -6503,7 +6324,7 @@ Exercise 2
-
+
R
@@ -6535,7 +6356,7 @@ Exercise 3
-
+
Whenever your code involves the generation of random numbers, it’s a good practice to set the random seed in R with @@ -791,7 +791,7 @@
Challenge
-R
@@ -1063,7 +1063,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1349,7 +1349,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -1512,7 +1512,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
diff --git a/hca.html b/hca.html
index 18576b9..071b6cc 100644
--- a/hca.html
+++ b/hca.html
@@ -788,7 +788,7 @@
Exercise 1
-
+
R
@@ -821,7 +821,7 @@ Exercise 2
-
+
R
@@ -853,7 +853,7 @@ Exercise 3
-
+
R
@@ -894,7 +894,7 @@ Exercise 4
-
+
R
diff --git a/index.html b/index.html
index 24dd0f3..6dfcdaf 100644
--- a/index.html
+++ b/index.html
@@ -372,7 +372,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -398,7 +398,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN
website.
@@ -429,7 +429,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -453,7 +453,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN website.
- Select the
.pkg file for the latest R version
@@ -486,7 +486,7 @@
Install R using your package manager and RStudio
-
+
- Follow the instructions for your distribution from CRAN, they provide
information to get the most recent version of R for common
diff --git a/instructor/aio.html b/instructor/aio.html
index 4714d2d..f7a8835 100644
--- a/instructor/aio.html
+++ b/instructor/aio.html
@@ -1040,7 +1040,7 @@
Challenge
-
+
R
@@ -1120,7 +1120,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -1353,7 +1353,7 @@
Challenge
-
+
R
@@ -1629,7 +1629,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1924,7 +1924,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -2094,7 +2094,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
@@ -2380,7 +2380,7 @@
Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -2533,7 +2533,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -2826,7 +2826,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -3015,7 +3015,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -3024,231 +3026,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -3292,7 +3086,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -3441,7 +3235,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -3458,7 +3252,7 @@
Give me a hint
-
+
TODO
@@ -3487,7 +3281,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -3501,7 +3295,7 @@
Give me a hint
-
+
TODO
@@ -3531,7 +3325,7 @@ Exercise 3: Workflow
-
+
TODO
@@ -3603,7 +3397,7 @@ Key Points
-->
Content from Multi-sample analyses
-Last updated on 2024-09-06 |
+
Last updated on 2024-09-08 |
Edit this page
Estimated time: 45 minutes
@@ -3661,13 +3455,23 @@ Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.
We start by loading the data and doing a quick exploratory analysis,
essentially applying the normalization and visualization techniques that
-we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this
+time we’re selecting samples 5 to 10, not just 5 by itself.
R
library(MouseGastrulationData)
-sce <- WTChimeraData(samples=5:10, type = "processed")
+library(batchelor)
+library(edgeR)
+library(scater)
+library(ggplot2)
+library(scran)
+library(pheatmap)
+library(scuttle)
+
+sce <- WTChimeraData(samples = 5:10, type = "processed")
+
sce
@@ -3742,15 +3546,12 @@ OUTPUT<
R
-library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3758,23 +3559,24 @@ R
sce <- sce[,idx]
-We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps,
+and visualize them in a tSNE plot.
R
-# normalization
-sce <- logNormCounts(sce)
+sce <- logNormCounts(sce)
+
+dec <- modelGeneVar(sce, block = sce$sample)
-# identify highly variable genes
-dec <- modelGeneVar(sce, block=sce$sample)
chosen.hvgs <- dec$bio > 0
-# dimensionality reduction
sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000)
+
sce <- runTSNE(sce, dimred = "PCA")
sce$sample <- as.factor(sce$sample)
+
plotTSNE(sce, colour_by = "sample")
Challenge
-
+
Samples 5 and 6 were from the same “pool” of cells. Looking at the
documentation for the dataset under ?WTChimeraData we see
@@ -3822,7 +3624,7 @@
Show me the solution
first TSNE plot. If you were developing single-cell library preparation
protocols you might want to preserve this effect to understand how
variation in pools leads to variation in expression, but for now, given
-that we’re investigate other effects, we’ll want to remove this as
+that we’re investigating other effects, we’ll want to remove this as
undesired technical variation.
@@ -3831,29 +3633,30 @@ Show me the solution
Correcting batch effects
-We correct the effect of samples by aid of the
+
We “correct” the effect of samples with the
correctExperiment function in the batchelor
-package and using the sample colData column as
-batch.
+package and using the sample column as batch.
R
-library(batchelor)
-set.seed(10102)
-merged <- correctExperiments(sce,
- batch=sce$sample,
- subset.row=chosen.hvgs,
- PARAM=FastMnnParam(
- merge.order=list(
+set.seed(10102)
+
+merged <- correctExperiments(
+ sce,
+ batch = sce$sample,
+ subset.row = chosen.hvgs,
+ PARAM = FastMnnParam(
+ merge.order = list(
list(1,3,5), # WT (3 replicates)
list(2,4,6) # td-Tomato (3 replicates)
)
)
)
-merged <- runTSNE(merged, dimred="corrected")
-plotTSNE(merged, colour_by="batch")
+merged <- runTSNE(merged, dimred = "corrected")
+
+plotTSNE(merged, colour_by = "batch")
Once we removed the sample batch effect, we can proceed with the
Differential Expression Analysis.
@@ -3875,7 +3678,7 @@ Challenge
-
+
It’s important to have multiple samples within each experimental
group because it helps the batch effect correction algorithm distinguish
@@ -3898,7 +3701,7 @@
Show me the solution
Differential Expression
-In order to perform a Differential Expression Analysis, we need to
+
In order to perform a differential expression analysis, we need to
identify groups of cells across samples/conditions (depending on the
experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell
@@ -3926,9 +3729,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output
# corresponds to one unique combination of these two factors.
-library(scuttle)
-summed <- aggregateAcrossCells(merged,
- id=colData(merged)[,c("celltype.mapped", "sample")])
+
+summed <- aggregateAcrossCells(
+ merged,
+ id = colData(merged)[,c("celltype.mapped", "sample")]
+)
+
summed
R
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
y
@@ -4018,7 +3824,9 @@ R
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)
@@ -4034,8 +3842,10 @@ OUTPUT<
R
-keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)
@@ -4055,6 +3865,7 @@ R
y <- calcNormFactors(y)
+
y$samples
@@ -4095,7 +3906,7 @@ Rpar(mfrow = c(2,3))
for (i in seq_len(ncol(y))) {
- plotMD(y, column=i)
+ plotMD(y, column = i)
}
RR
-limma::plotMDS(cpm(y, log=TRUE),
- col=ifelse(y$samples$tomato, "red", "blue"))
+limma::plotMDS(cpm(y, log = TRUE),
+ col = ifelse(y$samples$tomato, "red", "blue"))
We then construct a design matrix by including both the pool and the
tomato as factors. This design indicates which samples belong to which
@@ -4153,6 +3964,7 @@
R
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)
@@ -4172,7 +3984,7 @@ RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized
-linear model for each gene. The robust=TRUE parameter
+linear model for each gene. The robust = TRUE parameter
avoids distortions from highly variable clusters. The QL method includes
an additional dispersion parameter, useful to handle the uncertainty and
variability of the per-gene variance, which is not well estimated by the
@@ -4182,7 +3994,8 @@
RR
-fit <- glmQLFit(y, design, robust=TRUE)
+fit <- glmQLFit(y, design, robust = TRUE)
+
summary(fit$var.prior)
@@ -4221,7 +4034,8 @@ RR
-res <- glmQLFTest(fit, coef=ncol(design))
+res <- glmQLFTest(fit, coef = ncol(design))
+
summary(decideTests(res))
@@ -4261,14 +4075,14 @@ OUTPUT<
R
-library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)
The returned object is a list of DataFrames each with
@@ -4280,6 +4094,7 @@
R
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]
Challenge
-
+
“logFC” stands for log fold-change. Rather than reporting e.g. a
5-fold increase, it’s better to report a logFC of log(5) = 1.61.
@@ -4336,26 +4151,29 @@
Show me the solution
With DA we look for differences in cluster abundance across
conditions (the tomato injection in our case), rather than differences
in gene expression.
-We first setup some code and variables for further analysis, like
-quantifying the number of cells per each cell type and fit a model to
-catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this
-time we start our analysis on the computed abundances and without
+
Our first steps are quantifying the number of cells per each cell
+type and fitting a model to catch differences between the injected cells
+and the background.
+The process is very similar differential expression modeling, but
+this time we start our analysis on the computed abundances and without
normalizing the data with TMM.
R
-library(edgeR)
-abundances <- table(merged$celltype.mapped, merged$sample)
+abundances <- table(merged$celltype.mapped, merged$sample)
+
abundances <- unclass(abundances)
-# Attaching some column metadata.
+
extra.info <- colData(merged)[match(colnames(abundances), merged$sample),]
-y.ab <- DGEList(abundances, samples=extra.info)
+
+y.ab <- DGEList(abundances, samples = extra.info)
design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples)
-y.ab <- estimateDisp(y.ab, design, trend="none")
-fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)
+
+y.ab <- estimateDisp(y.ab, design, trend = "none")
+
+fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)
Background on compositional effect
@@ -4401,6 +4219,7 @@ R
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
@@ -4413,11 +4232,11 @@ OUTPUT<
R
-y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))
@@ -4433,7 +4252,7 @@ OUTPUT<
R
-topTags(res2, n=10)
+topTags(res2, n = 10)
OUTPUT
@@ -4469,7 +4288,8 @@ Testing against a log-fold
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
@@ -4550,70 +4370,72 @@ OUTPUT<
[8] base
other attached packages:
- [1] edgeR_4.2.0 limma_3.60.2
- [3] batchelor_1.20.0 scran_1.32.0
- [5] scater_1.32.0 ggplot2_3.5.1
- [7] scuttle_1.14.0 MouseGastrulationData_1.18.0
- [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0
-[11] SummarizedExperiment_1.34.0 Biobase_2.64.0
-[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1
-[15] IRanges_2.38.0 S4Vectors_0.42.0
-[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0
-[19] matrixStats_1.3.0 BiocStyle_2.32.0
+ [1] pheatmap_1.0.12 scran_1.32.0
+ [3] scater_1.32.0 ggplot2_3.5.1
+ [5] scuttle_1.14.0 edgeR_4.2.0
+ [7] limma_3.60.2 batchelor_1.20.0
+ [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0
+[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0
+[13] Biobase_2.64.0 GenomicRanges_1.56.0
+[15] GenomeInfoDb_1.40.1 IRanges_2.38.0
+[17] S4Vectors_0.42.0 BiocGenerics_0.50.0
+[19] MatrixGenerics_1.16.0 matrixStats_1.3.0
+[21] BiocStyle_2.32.0
loaded via a namespace (and not attached):
- [1] DBI_1.2.3 formatR_1.14
- [3] gridExtra_2.3 rlang_1.1.3
- [5] magrittr_2.0.3 compiler_4.4.1
- [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0
- [9] png_0.1-8 vctrs_0.6.5
- [11] pkgconfig_2.0.3 crayon_1.5.2
- [13] fastmap_1.2.0 dbplyr_2.5.0
- [15] magick_2.8.3 XVector_0.44.0
- [17] labeling_0.4.3 utf8_1.2.4
- [19] rmarkdown_2.27 UCSC.utils_1.0.0
- [21] ggbeeswarm_0.7.2 purrr_1.0.2
- [23] bit_4.0.5 bluster_1.14.0
- [25] xfun_0.44 zlibbioc_1.50.0
- [27] cachem_1.1.0 beachmat_2.20.0
- [29] jsonlite_1.8.8 blob_1.2.4
- [31] highr_0.11 DelayedArray_0.30.1
- [33] BiocParallel_1.38.0 cluster_2.1.6
- [35] irlba_2.3.5.1 parallel_4.4.1
- [37] R6_2.5.1 Rcpp_1.0.12
- [39] knitr_1.47 splines_4.4.1
- [41] igraph_2.0.3 Matrix_1.7-0
- [43] tidyselect_1.2.1 viridis_0.6.5
- [45] abind_1.4-5 yaml_2.3.8
- [47] codetools_0.2-20 curl_5.2.1
- [49] lattice_0.22-6 tibble_3.2.1
- [51] withr_3.0.0 KEGGREST_1.44.0
- [53] BumpyMatrix_1.12.0 Rtsne_0.17
- [55] evaluate_0.23 BiocFileCache_2.12.0
- [57] ExperimentHub_2.12.0 Biostrings_2.72.1
- [59] pillar_1.9.0 BiocManager_1.30.23
- [61] filelock_1.0.3 renv_1.0.7
- [63] generics_0.1.3 BiocVersion_3.19.1
- [65] sparseMatrixStats_1.16.0 munsell_0.5.1
- [67] scales_1.3.0 glue_1.7.0
- [69] metapod_1.12.0 tools_4.4.1
- [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0
- [73] ScaledMatrix_1.12.0 locfit_1.5-9.9
- [75] cowplot_1.1.3 grid_4.4.1
- [77] AnnotationDbi_1.66.0 colorspace_2.1-0
- [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0
- [81] BiocSingular_1.20.0 vipor_0.4.7
- [83] cli_3.6.2 rsvd_1.0.5
- [85] rappdirs_0.3.3 fansi_1.0.6
- [87] viridisLite_0.4.2 S4Arrays_1.4.1
- [89] dplyr_1.1.4 ResidualMatrix_1.14.0
- [91] gtable_0.3.5 digest_0.6.35
- [93] dqrng_0.4.1 SparseArray_1.4.8
- [95] ggrepel_0.9.5 farver_2.1.2
- [97] rjson_0.2.21 memoise_2.0.1
- [99] htmltools_0.5.8.1 lifecycle_1.0.4
-[101] httr_1.4.7 statmod_1.5.0
-[103] mime_0.12 bit64_4.0.5
+ [1] RColorBrewer_1.1-3 jsonlite_1.8.8
+ [3] magrittr_2.0.3 ggbeeswarm_0.7.2
+ [5] magick_2.8.3 farver_2.1.2
+ [7] rmarkdown_2.27 zlibbioc_1.50.0
+ [9] vctrs_0.6.5 memoise_2.0.1
+ [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1
+ [13] S4Arrays_1.4.1 AnnotationHub_3.12.0
+ [15] curl_5.2.1 BiocNeighbors_1.22.0
+ [17] SparseArray_1.4.8 cachem_1.1.0
+ [19] ResidualMatrix_1.14.0 igraph_2.0.3
+ [21] mime_0.12 lifecycle_1.0.4
+ [23] pkgconfig_2.0.3 rsvd_1.0.5
+ [25] Matrix_1.7-0 R6_2.5.1
+ [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12
+ [29] digest_0.6.35 colorspace_2.1-0
+ [31] AnnotationDbi_1.66.0 dqrng_0.4.1
+ [33] irlba_2.3.5.1 ExperimentHub_2.12.0
+ [35] RSQLite_2.3.7 beachmat_2.20.0
+ [37] filelock_1.0.3 labeling_0.4.3
+ [39] fansi_1.0.6 httr_1.4.7
+ [41] abind_1.4-5 compiler_4.4.1
+ [43] bit64_4.0.5 withr_3.0.0
+ [45] BiocParallel_1.38.0 viridis_0.6.5
+ [47] DBI_1.2.3 highr_0.11
+ [49] rappdirs_0.3.3 DelayedArray_0.30.1
+ [51] rjson_0.2.21 bluster_1.14.0
+ [53] tools_4.4.1 vipor_0.4.7
+ [55] beeswarm_0.4.0 glue_1.7.0
+ [57] grid_4.4.1 Rtsne_0.17
+ [59] cluster_2.1.6 generics_0.1.3
+ [61] gtable_0.3.5 BiocSingular_1.20.0
+ [63] ScaledMatrix_1.12.0 metapod_1.12.0
+ [65] utf8_1.2.4 XVector_0.44.0
+ [67] ggrepel_0.9.5 BiocVersion_3.19.1
+ [69] pillar_1.9.0 BumpyMatrix_1.12.0
+ [71] splines_4.4.1 dplyr_1.1.4
+ [73] BiocFileCache_2.12.0 lattice_0.22-6
+ [75] renv_1.0.7 bit_4.0.5
+ [77] tidyselect_1.2.1 locfit_1.5-9.9
+ [79] Biostrings_2.72.1 knitr_1.47
+ [81] gridExtra_2.3 xfun_0.44
+ [83] statmod_1.5.0 UCSC.utils_1.0.0
+ [85] yaml_2.3.8 evaluate_0.23
+ [87] codetools_0.2-20 tibble_3.2.1
+ [89] BiocManager_1.30.23 cli_3.6.2
+ [91] munsell_0.5.1 Rcpp_1.0.12
+ [93] dbplyr_2.5.0 png_0.1-8
+ [95] parallel_4.4.1 blob_1.2.4
+ [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2
+ [99] scales_1.3.0 purrr_1.0.2
+[101] crayon_1.5.2 rlang_1.1.3
+[103] cowplot_1.1.3 KEGGREST_1.44.0
+[105] formatR_1.14
Exercises
@@ -4636,7 +4458,7 @@ Exercise 1: Replicates
-
+
Remember, you can subset SingleCellExperiments with logical indices,
just like a matrix. You can also access their column data with the
@@ -4650,7 +4472,7 @@
Give me a hint
-
+
R
@@ -4658,11 +4480,12 @@ R
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
@@ -4686,7 +4509,7 @@ Exercise 2:
-
+
You can just hand pheatmap() a matrix as its only
argument. It has a million options, but the defaults are usually pretty
@@ -4700,15 +4523,13 @@
Give me a hint
-
+
R
-library(pheatmap)
-
-pheatmap(y.ab$counts)
+pheatmap(y.ab$counts)
The top DA result was a decrease in ExE ectoderm in the tomato
condition, which you can sort of see, especially if you
@@ -6470,7 +6291,7 @@
Exercise 1
-
+
R
@@ -6503,7 +6324,7 @@ Exercise 2
-
+
R
@@ -6535,7 +6356,7 @@ Exercise 3
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic or synthetic source at a known concentration. This provides a known @@ -1349,7 +1349,7 @@
Challenge
-Mathematically, this would require the data to fall on a two-dimensional plane (for linear methods like PCA) or a smooth 2D @@ -1512,7 +1512,7 @@
Exercise 2: Dimensionality Reduction
-First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
diff --git a/hca.html b/hca.html
index 18576b9..071b6cc 100644
--- a/hca.html
+++ b/hca.html
@@ -788,7 +788,7 @@
Exercise 1
-R
@@ -821,7 +821,7 @@ Exercise 2
-
+
R
@@ -853,7 +853,7 @@ Exercise 3
-
+
R
@@ -894,7 +894,7 @@ Exercise 4
-
+
R
diff --git a/index.html b/index.html
index 24dd0f3..6dfcdaf 100644
--- a/index.html
+++ b/index.html
@@ -372,7 +372,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -398,7 +398,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN
website.
@@ -429,7 +429,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -453,7 +453,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN website.
- Select the
.pkg file for the latest R version
@@ -486,7 +486,7 @@
Install R using your package manager and RStudio
-
+
- Follow the instructions for your distribution from CRAN, they provide
information to get the most recent version of R for common
diff --git a/instructor/aio.html b/instructor/aio.html
index 4714d2d..f7a8835 100644
--- a/instructor/aio.html
+++ b/instructor/aio.html
@@ -1040,7 +1040,7 @@
Challenge
-
+
R
@@ -1120,7 +1120,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -1353,7 +1353,7 @@
Challenge
-
+
R
@@ -1629,7 +1629,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1924,7 +1924,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -2094,7 +2094,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
@@ -2380,7 +2380,7 @@
Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -2533,7 +2533,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -2826,7 +2826,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -3015,7 +3015,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -3024,231 +3026,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -3292,7 +3086,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -3441,7 +3235,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -3458,7 +3252,7 @@
Give me a hint
-
+
TODO
@@ -3487,7 +3281,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -3501,7 +3295,7 @@
Give me a hint
-
+
TODO
@@ -3531,7 +3325,7 @@ Exercise 3: Workflow
-
+
TODO
@@ -3603,7 +3397,7 @@ Key Points
-->
Content from Multi-sample analyses
-Last updated on 2024-09-06 |
+
Last updated on 2024-09-08 |
Edit this page
Estimated time: 45 minutes
@@ -3661,13 +3455,23 @@ Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.
We start by loading the data and doing a quick exploratory analysis,
essentially applying the normalization and visualization techniques that
-we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this
+time we’re selecting samples 5 to 10, not just 5 by itself.
R
library(MouseGastrulationData)
-sce <- WTChimeraData(samples=5:10, type = "processed")
+library(batchelor)
+library(edgeR)
+library(scater)
+library(ggplot2)
+library(scran)
+library(pheatmap)
+library(scuttle)
+
+sce <- WTChimeraData(samples = 5:10, type = "processed")
+
sce
@@ -3742,15 +3546,12 @@ OUTPUT<
R
-library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3758,23 +3559,24 @@ R
sce <- sce[,idx]
-We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps,
+and visualize them in a tSNE plot.
R
-# normalization
-sce <- logNormCounts(sce)
+sce <- logNormCounts(sce)
+
+dec <- modelGeneVar(sce, block = sce$sample)
-# identify highly variable genes
-dec <- modelGeneVar(sce, block=sce$sample)
chosen.hvgs <- dec$bio > 0
-# dimensionality reduction
sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000)
+
sce <- runTSNE(sce, dimred = "PCA")
sce$sample <- as.factor(sce$sample)
+
plotTSNE(sce, colour_by = "sample")
Challenge
-
+
Samples 5 and 6 were from the same “pool” of cells. Looking at the
documentation for the dataset under ?WTChimeraData we see
@@ -3822,7 +3624,7 @@
Show me the solution
first TSNE plot. If you were developing single-cell library preparation
protocols you might want to preserve this effect to understand how
variation in pools leads to variation in expression, but for now, given
-that we’re investigate other effects, we’ll want to remove this as
+that we’re investigating other effects, we’ll want to remove this as
undesired technical variation.
@@ -3831,29 +3633,30 @@ Show me the solution
Correcting batch effects
-We correct the effect of samples by aid of the
+
We “correct” the effect of samples with the
correctExperiment function in the batchelor
-package and using the sample colData column as
-batch.
+package and using the sample column as batch.
R
-library(batchelor)
-set.seed(10102)
-merged <- correctExperiments(sce,
- batch=sce$sample,
- subset.row=chosen.hvgs,
- PARAM=FastMnnParam(
- merge.order=list(
+set.seed(10102)
+
+merged <- correctExperiments(
+ sce,
+ batch = sce$sample,
+ subset.row = chosen.hvgs,
+ PARAM = FastMnnParam(
+ merge.order = list(
list(1,3,5), # WT (3 replicates)
list(2,4,6) # td-Tomato (3 replicates)
)
)
)
-merged <- runTSNE(merged, dimred="corrected")
-plotTSNE(merged, colour_by="batch")
+merged <- runTSNE(merged, dimred = "corrected")
+
+plotTSNE(merged, colour_by = "batch")
Once we removed the sample batch effect, we can proceed with the
Differential Expression Analysis.
@@ -3875,7 +3678,7 @@ Challenge
-
+
It’s important to have multiple samples within each experimental
group because it helps the batch effect correction algorithm distinguish
@@ -3898,7 +3701,7 @@
Show me the solution
Differential Expression
-In order to perform a Differential Expression Analysis, we need to
+
In order to perform a differential expression analysis, we need to
identify groups of cells across samples/conditions (depending on the
experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell
@@ -3926,9 +3729,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output
# corresponds to one unique combination of these two factors.
-library(scuttle)
-summed <- aggregateAcrossCells(merged,
- id=colData(merged)[,c("celltype.mapped", "sample")])
+
+summed <- aggregateAcrossCells(
+ merged,
+ id = colData(merged)[,c("celltype.mapped", "sample")]
+)
+
summed
R
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
y
@@ -4018,7 +3824,9 @@ R
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)
@@ -4034,8 +3842,10 @@ OUTPUT<
R
-keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)
@@ -4055,6 +3865,7 @@ R
y <- calcNormFactors(y)
+
y$samples
@@ -4095,7 +3906,7 @@ Rpar(mfrow = c(2,3))
for (i in seq_len(ncol(y))) {
- plotMD(y, column=i)
+ plotMD(y, column = i)
}
RR
-limma::plotMDS(cpm(y, log=TRUE),
- col=ifelse(y$samples$tomato, "red", "blue"))
+limma::plotMDS(cpm(y, log = TRUE),
+ col = ifelse(y$samples$tomato, "red", "blue"))
We then construct a design matrix by including both the pool and the
tomato as factors. This design indicates which samples belong to which
@@ -4153,6 +3964,7 @@
R
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)
@@ -4172,7 +3984,7 @@ RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized
-linear model for each gene. The robust=TRUE parameter
+linear model for each gene. The robust = TRUE parameter
avoids distortions from highly variable clusters. The QL method includes
an additional dispersion parameter, useful to handle the uncertainty and
variability of the per-gene variance, which is not well estimated by the
@@ -4182,7 +3994,8 @@
RR
-fit <- glmQLFit(y, design, robust=TRUE)
+fit <- glmQLFit(y, design, robust = TRUE)
+
summary(fit$var.prior)
@@ -4221,7 +4034,8 @@ RR
-res <- glmQLFTest(fit, coef=ncol(design))
+res <- glmQLFTest(fit, coef = ncol(design))
+
summary(decideTests(res))
@@ -4261,14 +4075,14 @@ OUTPUT<
R
-library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)
The returned object is a list of DataFrames each with
@@ -4280,6 +4094,7 @@
R
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]
Challenge
-
+
“logFC” stands for log fold-change. Rather than reporting e.g. a
5-fold increase, it’s better to report a logFC of log(5) = 1.61.
@@ -4336,26 +4151,29 @@
Show me the solution
With DA we look for differences in cluster abundance across
conditions (the tomato injection in our case), rather than differences
in gene expression.
-We first setup some code and variables for further analysis, like
-quantifying the number of cells per each cell type and fit a model to
-catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this
-time we start our analysis on the computed abundances and without
+
Our first steps are quantifying the number of cells per each cell
+type and fitting a model to catch differences between the injected cells
+and the background.
+The process is very similar differential expression modeling, but
+this time we start our analysis on the computed abundances and without
normalizing the data with TMM.
R
-library(edgeR)
-abundances <- table(merged$celltype.mapped, merged$sample)
+abundances <- table(merged$celltype.mapped, merged$sample)
+
abundances <- unclass(abundances)
-# Attaching some column metadata.
+
extra.info <- colData(merged)[match(colnames(abundances), merged$sample),]
-y.ab <- DGEList(abundances, samples=extra.info)
+
+y.ab <- DGEList(abundances, samples = extra.info)
design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples)
-y.ab <- estimateDisp(y.ab, design, trend="none")
-fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)
+
+y.ab <- estimateDisp(y.ab, design, trend = "none")
+
+fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)
Background on compositional effect
@@ -4401,6 +4219,7 @@ R
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
@@ -4413,11 +4232,11 @@ OUTPUT<
R
-y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))
@@ -4433,7 +4252,7 @@ OUTPUT<
R
-topTags(res2, n=10)
+topTags(res2, n = 10)
OUTPUT
@@ -4469,7 +4288,8 @@ Testing against a log-fold
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
@@ -4550,70 +4370,72 @@ OUTPUT<
[8] base
other attached packages:
- [1] edgeR_4.2.0 limma_3.60.2
- [3] batchelor_1.20.0 scran_1.32.0
- [5] scater_1.32.0 ggplot2_3.5.1
- [7] scuttle_1.14.0 MouseGastrulationData_1.18.0
- [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0
-[11] SummarizedExperiment_1.34.0 Biobase_2.64.0
-[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1
-[15] IRanges_2.38.0 S4Vectors_0.42.0
-[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0
-[19] matrixStats_1.3.0 BiocStyle_2.32.0
+ [1] pheatmap_1.0.12 scran_1.32.0
+ [3] scater_1.32.0 ggplot2_3.5.1
+ [5] scuttle_1.14.0 edgeR_4.2.0
+ [7] limma_3.60.2 batchelor_1.20.0
+ [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0
+[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0
+[13] Biobase_2.64.0 GenomicRanges_1.56.0
+[15] GenomeInfoDb_1.40.1 IRanges_2.38.0
+[17] S4Vectors_0.42.0 BiocGenerics_0.50.0
+[19] MatrixGenerics_1.16.0 matrixStats_1.3.0
+[21] BiocStyle_2.32.0
loaded via a namespace (and not attached):
- [1] DBI_1.2.3 formatR_1.14
- [3] gridExtra_2.3 rlang_1.1.3
- [5] magrittr_2.0.3 compiler_4.4.1
- [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0
- [9] png_0.1-8 vctrs_0.6.5
- [11] pkgconfig_2.0.3 crayon_1.5.2
- [13] fastmap_1.2.0 dbplyr_2.5.0
- [15] magick_2.8.3 XVector_0.44.0
- [17] labeling_0.4.3 utf8_1.2.4
- [19] rmarkdown_2.27 UCSC.utils_1.0.0
- [21] ggbeeswarm_0.7.2 purrr_1.0.2
- [23] bit_4.0.5 bluster_1.14.0
- [25] xfun_0.44 zlibbioc_1.50.0
- [27] cachem_1.1.0 beachmat_2.20.0
- [29] jsonlite_1.8.8 blob_1.2.4
- [31] highr_0.11 DelayedArray_0.30.1
- [33] BiocParallel_1.38.0 cluster_2.1.6
- [35] irlba_2.3.5.1 parallel_4.4.1
- [37] R6_2.5.1 Rcpp_1.0.12
- [39] knitr_1.47 splines_4.4.1
- [41] igraph_2.0.3 Matrix_1.7-0
- [43] tidyselect_1.2.1 viridis_0.6.5
- [45] abind_1.4-5 yaml_2.3.8
- [47] codetools_0.2-20 curl_5.2.1
- [49] lattice_0.22-6 tibble_3.2.1
- [51] withr_3.0.0 KEGGREST_1.44.0
- [53] BumpyMatrix_1.12.0 Rtsne_0.17
- [55] evaluate_0.23 BiocFileCache_2.12.0
- [57] ExperimentHub_2.12.0 Biostrings_2.72.1
- [59] pillar_1.9.0 BiocManager_1.30.23
- [61] filelock_1.0.3 renv_1.0.7
- [63] generics_0.1.3 BiocVersion_3.19.1
- [65] sparseMatrixStats_1.16.0 munsell_0.5.1
- [67] scales_1.3.0 glue_1.7.0
- [69] metapod_1.12.0 tools_4.4.1
- [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0
- [73] ScaledMatrix_1.12.0 locfit_1.5-9.9
- [75] cowplot_1.1.3 grid_4.4.1
- [77] AnnotationDbi_1.66.0 colorspace_2.1-0
- [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0
- [81] BiocSingular_1.20.0 vipor_0.4.7
- [83] cli_3.6.2 rsvd_1.0.5
- [85] rappdirs_0.3.3 fansi_1.0.6
- [87] viridisLite_0.4.2 S4Arrays_1.4.1
- [89] dplyr_1.1.4 ResidualMatrix_1.14.0
- [91] gtable_0.3.5 digest_0.6.35
- [93] dqrng_0.4.1 SparseArray_1.4.8
- [95] ggrepel_0.9.5 farver_2.1.2
- [97] rjson_0.2.21 memoise_2.0.1
- [99] htmltools_0.5.8.1 lifecycle_1.0.4
-[101] httr_1.4.7 statmod_1.5.0
-[103] mime_0.12 bit64_4.0.5
+ [1] RColorBrewer_1.1-3 jsonlite_1.8.8
+ [3] magrittr_2.0.3 ggbeeswarm_0.7.2
+ [5] magick_2.8.3 farver_2.1.2
+ [7] rmarkdown_2.27 zlibbioc_1.50.0
+ [9] vctrs_0.6.5 memoise_2.0.1
+ [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1
+ [13] S4Arrays_1.4.1 AnnotationHub_3.12.0
+ [15] curl_5.2.1 BiocNeighbors_1.22.0
+ [17] SparseArray_1.4.8 cachem_1.1.0
+ [19] ResidualMatrix_1.14.0 igraph_2.0.3
+ [21] mime_0.12 lifecycle_1.0.4
+ [23] pkgconfig_2.0.3 rsvd_1.0.5
+ [25] Matrix_1.7-0 R6_2.5.1
+ [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12
+ [29] digest_0.6.35 colorspace_2.1-0
+ [31] AnnotationDbi_1.66.0 dqrng_0.4.1
+ [33] irlba_2.3.5.1 ExperimentHub_2.12.0
+ [35] RSQLite_2.3.7 beachmat_2.20.0
+ [37] filelock_1.0.3 labeling_0.4.3
+ [39] fansi_1.0.6 httr_1.4.7
+ [41] abind_1.4-5 compiler_4.4.1
+ [43] bit64_4.0.5 withr_3.0.0
+ [45] BiocParallel_1.38.0 viridis_0.6.5
+ [47] DBI_1.2.3 highr_0.11
+ [49] rappdirs_0.3.3 DelayedArray_0.30.1
+ [51] rjson_0.2.21 bluster_1.14.0
+ [53] tools_4.4.1 vipor_0.4.7
+ [55] beeswarm_0.4.0 glue_1.7.0
+ [57] grid_4.4.1 Rtsne_0.17
+ [59] cluster_2.1.6 generics_0.1.3
+ [61] gtable_0.3.5 BiocSingular_1.20.0
+ [63] ScaledMatrix_1.12.0 metapod_1.12.0
+ [65] utf8_1.2.4 XVector_0.44.0
+ [67] ggrepel_0.9.5 BiocVersion_3.19.1
+ [69] pillar_1.9.0 BumpyMatrix_1.12.0
+ [71] splines_4.4.1 dplyr_1.1.4
+ [73] BiocFileCache_2.12.0 lattice_0.22-6
+ [75] renv_1.0.7 bit_4.0.5
+ [77] tidyselect_1.2.1 locfit_1.5-9.9
+ [79] Biostrings_2.72.1 knitr_1.47
+ [81] gridExtra_2.3 xfun_0.44
+ [83] statmod_1.5.0 UCSC.utils_1.0.0
+ [85] yaml_2.3.8 evaluate_0.23
+ [87] codetools_0.2-20 tibble_3.2.1
+ [89] BiocManager_1.30.23 cli_3.6.2
+ [91] munsell_0.5.1 Rcpp_1.0.12
+ [93] dbplyr_2.5.0 png_0.1-8
+ [95] parallel_4.4.1 blob_1.2.4
+ [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2
+ [99] scales_1.3.0 purrr_1.0.2
+[101] crayon_1.5.2 rlang_1.1.3
+[103] cowplot_1.1.3 KEGGREST_1.44.0
+[105] formatR_1.14
Exercises
@@ -4636,7 +4458,7 @@ Exercise 1: Replicates
-
+
Remember, you can subset SingleCellExperiments with logical indices,
just like a matrix. You can also access their column data with the
@@ -4650,7 +4472,7 @@
Give me a hint
-
+
R
@@ -4658,11 +4480,12 @@ R
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
@@ -4686,7 +4509,7 @@ Exercise 2:
-
+
You can just hand pheatmap() a matrix as its only
argument. It has a million options, but the defaults are usually pretty
@@ -4700,15 +4523,13 @@
Give me a hint
-
+
R
-library(pheatmap)
-
-pheatmap(y.ab$counts)
+pheatmap(y.ab$counts)
The top DA result was a decrease in ExE ectoderm in the tomato
condition, which you can sort of see, especially if you
@@ -6470,7 +6291,7 @@
Exercise 1
-
+
R
@@ -6503,7 +6324,7 @@ Exercise 2
-
+
R
@@ -6535,7 +6356,7 @@ Exercise 3
-
+
R
@@ -853,7 +853,7 @@ Exercise 3
-
+
R
@@ -894,7 +894,7 @@ Exercise 4
-
+
R
diff --git a/index.html b/index.html
index 24dd0f3..6dfcdaf 100644
--- a/index.html
+++ b/index.html
@@ -372,7 +372,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -398,7 +398,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN
website.
@@ -429,7 +429,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -453,7 +453,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN website.
- Select the
.pkg file for the latest R version
@@ -486,7 +486,7 @@
Install R using your package manager and RStudio
-
+
- Follow the instructions for your distribution from CRAN, they provide
information to get the most recent version of R for common
diff --git a/instructor/aio.html b/instructor/aio.html
index 4714d2d..f7a8835 100644
--- a/instructor/aio.html
+++ b/instructor/aio.html
@@ -1040,7 +1040,7 @@
Challenge
-
+
R
@@ -1120,7 +1120,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -1353,7 +1353,7 @@
Challenge
-
+
R
@@ -1629,7 +1629,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1924,7 +1924,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -2094,7 +2094,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
@@ -2380,7 +2380,7 @@
Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -2533,7 +2533,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -2826,7 +2826,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -3015,7 +3015,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -3024,231 +3026,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -3292,7 +3086,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -3441,7 +3235,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -3458,7 +3252,7 @@
Give me a hint
-
+
TODO
@@ -3487,7 +3281,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -3501,7 +3295,7 @@
Give me a hint
-
+
TODO
@@ -3531,7 +3325,7 @@ Exercise 3: Workflow
-
+
TODO
@@ -3603,7 +3397,7 @@ Key Points
-->
Content from Multi-sample analyses
-Last updated on 2024-09-06 |
+
Last updated on 2024-09-08 |
Edit this page
Estimated time: 45 minutes
@@ -3661,13 +3455,23 @@ Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.
We start by loading the data and doing a quick exploratory analysis,
essentially applying the normalization and visualization techniques that
-we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this
+time we’re selecting samples 5 to 10, not just 5 by itself.
R
library(MouseGastrulationData)
-sce <- WTChimeraData(samples=5:10, type = "processed")
+library(batchelor)
+library(edgeR)
+library(scater)
+library(ggplot2)
+library(scran)
+library(pheatmap)
+library(scuttle)
+
+sce <- WTChimeraData(samples = 5:10, type = "processed")
+
sce
@@ -3742,15 +3546,12 @@ OUTPUT<
R
-library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3758,23 +3559,24 @@ R
sce <- sce[,idx]
-We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps,
+and visualize them in a tSNE plot.
R
-# normalization
-sce <- logNormCounts(sce)
+sce <- logNormCounts(sce)
+
+dec <- modelGeneVar(sce, block = sce$sample)
-# identify highly variable genes
-dec <- modelGeneVar(sce, block=sce$sample)
chosen.hvgs <- dec$bio > 0
-# dimensionality reduction
sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000)
+
sce <- runTSNE(sce, dimred = "PCA")
sce$sample <- as.factor(sce$sample)
+
plotTSNE(sce, colour_by = "sample")
Challenge
-
+
Samples 5 and 6 were from the same “pool” of cells. Looking at the
documentation for the dataset under ?WTChimeraData we see
@@ -3822,7 +3624,7 @@
Show me the solution
first TSNE plot. If you were developing single-cell library preparation
protocols you might want to preserve this effect to understand how
variation in pools leads to variation in expression, but for now, given
-that we’re investigate other effects, we’ll want to remove this as
+that we’re investigating other effects, we’ll want to remove this as
undesired technical variation.
@@ -3831,29 +3633,30 @@ Show me the solution
Correcting batch effects
-We correct the effect of samples by aid of the
+
We “correct” the effect of samples with the
correctExperiment function in the batchelor
-package and using the sample colData column as
-batch.
+package and using the sample column as batch.
R
-library(batchelor)
-set.seed(10102)
-merged <- correctExperiments(sce,
- batch=sce$sample,
- subset.row=chosen.hvgs,
- PARAM=FastMnnParam(
- merge.order=list(
+set.seed(10102)
+
+merged <- correctExperiments(
+ sce,
+ batch = sce$sample,
+ subset.row = chosen.hvgs,
+ PARAM = FastMnnParam(
+ merge.order = list(
list(1,3,5), # WT (3 replicates)
list(2,4,6) # td-Tomato (3 replicates)
)
)
)
-merged <- runTSNE(merged, dimred="corrected")
-plotTSNE(merged, colour_by="batch")
+merged <- runTSNE(merged, dimred = "corrected")
+
+plotTSNE(merged, colour_by = "batch")
Once we removed the sample batch effect, we can proceed with the
Differential Expression Analysis.
@@ -3875,7 +3678,7 @@ Challenge
-
+
It’s important to have multiple samples within each experimental
group because it helps the batch effect correction algorithm distinguish
@@ -3898,7 +3701,7 @@
Show me the solution
Differential Expression
-In order to perform a Differential Expression Analysis, we need to
+
In order to perform a differential expression analysis, we need to
identify groups of cells across samples/conditions (depending on the
experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell
@@ -3926,9 +3729,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output
# corresponds to one unique combination of these two factors.
-library(scuttle)
-summed <- aggregateAcrossCells(merged,
- id=colData(merged)[,c("celltype.mapped", "sample")])
+
+summed <- aggregateAcrossCells(
+ merged,
+ id = colData(merged)[,c("celltype.mapped", "sample")]
+)
+
summed
R
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
y
@@ -4018,7 +3824,9 @@ R
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)
@@ -4034,8 +3842,10 @@ OUTPUT<
R
-keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)
@@ -4055,6 +3865,7 @@ R
y <- calcNormFactors(y)
+
y$samples
@@ -4095,7 +3906,7 @@ Rpar(mfrow = c(2,3))
for (i in seq_len(ncol(y))) {
- plotMD(y, column=i)
+ plotMD(y, column = i)
}
RR
-limma::plotMDS(cpm(y, log=TRUE),
- col=ifelse(y$samples$tomato, "red", "blue"))
+limma::plotMDS(cpm(y, log = TRUE),
+ col = ifelse(y$samples$tomato, "red", "blue"))
We then construct a design matrix by including both the pool and the
tomato as factors. This design indicates which samples belong to which
@@ -4153,6 +3964,7 @@
R
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)
@@ -4172,7 +3984,7 @@ RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized
-linear model for each gene. The robust=TRUE parameter
+linear model for each gene. The robust = TRUE parameter
avoids distortions from highly variable clusters. The QL method includes
an additional dispersion parameter, useful to handle the uncertainty and
variability of the per-gene variance, which is not well estimated by the
@@ -4182,7 +3994,8 @@
RR
-fit <- glmQLFit(y, design, robust=TRUE)
+fit <- glmQLFit(y, design, robust = TRUE)
+
summary(fit$var.prior)
@@ -4221,7 +4034,8 @@ RR
-res <- glmQLFTest(fit, coef=ncol(design))
+res <- glmQLFTest(fit, coef = ncol(design))
+
summary(decideTests(res))
@@ -4261,14 +4075,14 @@ OUTPUT<
R
-library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)
The returned object is a list of DataFrames each with
@@ -4280,6 +4094,7 @@
R
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]
Challenge
-
+
“logFC” stands for log fold-change. Rather than reporting e.g. a
5-fold increase, it’s better to report a logFC of log(5) = 1.61.
@@ -4336,26 +4151,29 @@
Show me the solution
With DA we look for differences in cluster abundance across
conditions (the tomato injection in our case), rather than differences
in gene expression.
-We first setup some code and variables for further analysis, like
-quantifying the number of cells per each cell type and fit a model to
-catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this
-time we start our analysis on the computed abundances and without
+
Our first steps are quantifying the number of cells per each cell
+type and fitting a model to catch differences between the injected cells
+and the background.
+The process is very similar differential expression modeling, but
+this time we start our analysis on the computed abundances and without
normalizing the data with TMM.
R
-library(edgeR)
-abundances <- table(merged$celltype.mapped, merged$sample)
+abundances <- table(merged$celltype.mapped, merged$sample)
+
abundances <- unclass(abundances)
-# Attaching some column metadata.
+
extra.info <- colData(merged)[match(colnames(abundances), merged$sample),]
-y.ab <- DGEList(abundances, samples=extra.info)
+
+y.ab <- DGEList(abundances, samples = extra.info)
design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples)
-y.ab <- estimateDisp(y.ab, design, trend="none")
-fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)
+
+y.ab <- estimateDisp(y.ab, design, trend = "none")
+
+fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)
Background on compositional effect
@@ -4401,6 +4219,7 @@ R
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
@@ -4413,11 +4232,11 @@ OUTPUT<
R
-y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))
@@ -4433,7 +4252,7 @@ OUTPUT<
R
-topTags(res2, n=10)
+topTags(res2, n = 10)
OUTPUT
@@ -4469,7 +4288,8 @@ Testing against a log-fold
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
@@ -4550,70 +4370,72 @@ OUTPUT<
[8] base
other attached packages:
- [1] edgeR_4.2.0 limma_3.60.2
- [3] batchelor_1.20.0 scran_1.32.0
- [5] scater_1.32.0 ggplot2_3.5.1
- [7] scuttle_1.14.0 MouseGastrulationData_1.18.0
- [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0
-[11] SummarizedExperiment_1.34.0 Biobase_2.64.0
-[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1
-[15] IRanges_2.38.0 S4Vectors_0.42.0
-[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0
-[19] matrixStats_1.3.0 BiocStyle_2.32.0
+ [1] pheatmap_1.0.12 scran_1.32.0
+ [3] scater_1.32.0 ggplot2_3.5.1
+ [5] scuttle_1.14.0 edgeR_4.2.0
+ [7] limma_3.60.2 batchelor_1.20.0
+ [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0
+[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0
+[13] Biobase_2.64.0 GenomicRanges_1.56.0
+[15] GenomeInfoDb_1.40.1 IRanges_2.38.0
+[17] S4Vectors_0.42.0 BiocGenerics_0.50.0
+[19] MatrixGenerics_1.16.0 matrixStats_1.3.0
+[21] BiocStyle_2.32.0
loaded via a namespace (and not attached):
- [1] DBI_1.2.3 formatR_1.14
- [3] gridExtra_2.3 rlang_1.1.3
- [5] magrittr_2.0.3 compiler_4.4.1
- [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0
- [9] png_0.1-8 vctrs_0.6.5
- [11] pkgconfig_2.0.3 crayon_1.5.2
- [13] fastmap_1.2.0 dbplyr_2.5.0
- [15] magick_2.8.3 XVector_0.44.0
- [17] labeling_0.4.3 utf8_1.2.4
- [19] rmarkdown_2.27 UCSC.utils_1.0.0
- [21] ggbeeswarm_0.7.2 purrr_1.0.2
- [23] bit_4.0.5 bluster_1.14.0
- [25] xfun_0.44 zlibbioc_1.50.0
- [27] cachem_1.1.0 beachmat_2.20.0
- [29] jsonlite_1.8.8 blob_1.2.4
- [31] highr_0.11 DelayedArray_0.30.1
- [33] BiocParallel_1.38.0 cluster_2.1.6
- [35] irlba_2.3.5.1 parallel_4.4.1
- [37] R6_2.5.1 Rcpp_1.0.12
- [39] knitr_1.47 splines_4.4.1
- [41] igraph_2.0.3 Matrix_1.7-0
- [43] tidyselect_1.2.1 viridis_0.6.5
- [45] abind_1.4-5 yaml_2.3.8
- [47] codetools_0.2-20 curl_5.2.1
- [49] lattice_0.22-6 tibble_3.2.1
- [51] withr_3.0.0 KEGGREST_1.44.0
- [53] BumpyMatrix_1.12.0 Rtsne_0.17
- [55] evaluate_0.23 BiocFileCache_2.12.0
- [57] ExperimentHub_2.12.0 Biostrings_2.72.1
- [59] pillar_1.9.0 BiocManager_1.30.23
- [61] filelock_1.0.3 renv_1.0.7
- [63] generics_0.1.3 BiocVersion_3.19.1
- [65] sparseMatrixStats_1.16.0 munsell_0.5.1
- [67] scales_1.3.0 glue_1.7.0
- [69] metapod_1.12.0 tools_4.4.1
- [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0
- [73] ScaledMatrix_1.12.0 locfit_1.5-9.9
- [75] cowplot_1.1.3 grid_4.4.1
- [77] AnnotationDbi_1.66.0 colorspace_2.1-0
- [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0
- [81] BiocSingular_1.20.0 vipor_0.4.7
- [83] cli_3.6.2 rsvd_1.0.5
- [85] rappdirs_0.3.3 fansi_1.0.6
- [87] viridisLite_0.4.2 S4Arrays_1.4.1
- [89] dplyr_1.1.4 ResidualMatrix_1.14.0
- [91] gtable_0.3.5 digest_0.6.35
- [93] dqrng_0.4.1 SparseArray_1.4.8
- [95] ggrepel_0.9.5 farver_2.1.2
- [97] rjson_0.2.21 memoise_2.0.1
- [99] htmltools_0.5.8.1 lifecycle_1.0.4
-[101] httr_1.4.7 statmod_1.5.0
-[103] mime_0.12 bit64_4.0.5
+ [1] RColorBrewer_1.1-3 jsonlite_1.8.8
+ [3] magrittr_2.0.3 ggbeeswarm_0.7.2
+ [5] magick_2.8.3 farver_2.1.2
+ [7] rmarkdown_2.27 zlibbioc_1.50.0
+ [9] vctrs_0.6.5 memoise_2.0.1
+ [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1
+ [13] S4Arrays_1.4.1 AnnotationHub_3.12.0
+ [15] curl_5.2.1 BiocNeighbors_1.22.0
+ [17] SparseArray_1.4.8 cachem_1.1.0
+ [19] ResidualMatrix_1.14.0 igraph_2.0.3
+ [21] mime_0.12 lifecycle_1.0.4
+ [23] pkgconfig_2.0.3 rsvd_1.0.5
+ [25] Matrix_1.7-0 R6_2.5.1
+ [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12
+ [29] digest_0.6.35 colorspace_2.1-0
+ [31] AnnotationDbi_1.66.0 dqrng_0.4.1
+ [33] irlba_2.3.5.1 ExperimentHub_2.12.0
+ [35] RSQLite_2.3.7 beachmat_2.20.0
+ [37] filelock_1.0.3 labeling_0.4.3
+ [39] fansi_1.0.6 httr_1.4.7
+ [41] abind_1.4-5 compiler_4.4.1
+ [43] bit64_4.0.5 withr_3.0.0
+ [45] BiocParallel_1.38.0 viridis_0.6.5
+ [47] DBI_1.2.3 highr_0.11
+ [49] rappdirs_0.3.3 DelayedArray_0.30.1
+ [51] rjson_0.2.21 bluster_1.14.0
+ [53] tools_4.4.1 vipor_0.4.7
+ [55] beeswarm_0.4.0 glue_1.7.0
+ [57] grid_4.4.1 Rtsne_0.17
+ [59] cluster_2.1.6 generics_0.1.3
+ [61] gtable_0.3.5 BiocSingular_1.20.0
+ [63] ScaledMatrix_1.12.0 metapod_1.12.0
+ [65] utf8_1.2.4 XVector_0.44.0
+ [67] ggrepel_0.9.5 BiocVersion_3.19.1
+ [69] pillar_1.9.0 BumpyMatrix_1.12.0
+ [71] splines_4.4.1 dplyr_1.1.4
+ [73] BiocFileCache_2.12.0 lattice_0.22-6
+ [75] renv_1.0.7 bit_4.0.5
+ [77] tidyselect_1.2.1 locfit_1.5-9.9
+ [79] Biostrings_2.72.1 knitr_1.47
+ [81] gridExtra_2.3 xfun_0.44
+ [83] statmod_1.5.0 UCSC.utils_1.0.0
+ [85] yaml_2.3.8 evaluate_0.23
+ [87] codetools_0.2-20 tibble_3.2.1
+ [89] BiocManager_1.30.23 cli_3.6.2
+ [91] munsell_0.5.1 Rcpp_1.0.12
+ [93] dbplyr_2.5.0 png_0.1-8
+ [95] parallel_4.4.1 blob_1.2.4
+ [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2
+ [99] scales_1.3.0 purrr_1.0.2
+[101] crayon_1.5.2 rlang_1.1.3
+[103] cowplot_1.1.3 KEGGREST_1.44.0
+[105] formatR_1.14
Exercises
@@ -4636,7 +4458,7 @@ Exercise 1: Replicates
-
+
Remember, you can subset SingleCellExperiments with logical indices,
just like a matrix. You can also access their column data with the
@@ -4650,7 +4472,7 @@
Give me a hint
-
+
R
@@ -4658,11 +4480,12 @@ R
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
@@ -4686,7 +4509,7 @@ Exercise 2:
-
+
You can just hand pheatmap() a matrix as its only
argument. It has a million options, but the defaults are usually pretty
@@ -4700,15 +4523,13 @@
Give me a hint
-
+
R
-library(pheatmap)
-
-pheatmap(y.ab$counts)
+pheatmap(y.ab$counts)
The top DA result was a decrease in ExE ectoderm in the tomato
condition, which you can sort of see, especially if you
@@ -6470,7 +6291,7 @@
Exercise 1
-
+
R
@@ -6503,7 +6324,7 @@ Exercise 2
-
+
R
@@ -6535,7 +6356,7 @@ Exercise 3
-
+
R
@@ -894,7 +894,7 @@ Exercise 4
-
+
R
diff --git a/index.html b/index.html
index 24dd0f3..6dfcdaf 100644
--- a/index.html
+++ b/index.html
@@ -372,7 +372,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -398,7 +398,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN
website.
@@ -429,7 +429,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -453,7 +453,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN website.
- Select the
.pkg file for the latest R version
@@ -486,7 +486,7 @@
Install R using your package manager and RStudio
-
+
- Follow the instructions for your distribution from CRAN, they provide
information to get the most recent version of R for common
diff --git a/instructor/aio.html b/instructor/aio.html
index 4714d2d..f7a8835 100644
--- a/instructor/aio.html
+++ b/instructor/aio.html
@@ -1040,7 +1040,7 @@
Challenge
-
+
R
@@ -1120,7 +1120,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -1353,7 +1353,7 @@
Challenge
-
+
R
@@ -1629,7 +1629,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1924,7 +1924,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -2094,7 +2094,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
@@ -2380,7 +2380,7 @@
Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -2533,7 +2533,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -2826,7 +2826,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -3015,7 +3015,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -3024,231 +3026,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -3292,7 +3086,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -3441,7 +3235,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -3458,7 +3252,7 @@
Give me a hint
-
+
TODO
@@ -3487,7 +3281,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -3501,7 +3295,7 @@
Give me a hint
-
+
TODO
@@ -3531,7 +3325,7 @@ Exercise 3: Workflow
-
+
TODO
@@ -3603,7 +3397,7 @@ Key Points
-->
Content from Multi-sample analyses
-Last updated on 2024-09-06 |
+
Last updated on 2024-09-08 |
Edit this page
Estimated time: 45 minutes
@@ -3661,13 +3455,23 @@ Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.
We start by loading the data and doing a quick exploratory analysis,
essentially applying the normalization and visualization techniques that
-we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this
+time we’re selecting samples 5 to 10, not just 5 by itself.
R
library(MouseGastrulationData)
-sce <- WTChimeraData(samples=5:10, type = "processed")
+library(batchelor)
+library(edgeR)
+library(scater)
+library(ggplot2)
+library(scran)
+library(pheatmap)
+library(scuttle)
+
+sce <- WTChimeraData(samples = 5:10, type = "processed")
+
sce
@@ -3742,15 +3546,12 @@ OUTPUT<
R
-library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3758,23 +3559,24 @@ R
sce <- sce[,idx]
-We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps,
+and visualize them in a tSNE plot.
R
-# normalization
-sce <- logNormCounts(sce)
+sce <- logNormCounts(sce)
+
+dec <- modelGeneVar(sce, block = sce$sample)
-# identify highly variable genes
-dec <- modelGeneVar(sce, block=sce$sample)
chosen.hvgs <- dec$bio > 0
-# dimensionality reduction
sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000)
+
sce <- runTSNE(sce, dimred = "PCA")
sce$sample <- as.factor(sce$sample)
+
plotTSNE(sce, colour_by = "sample")
Challenge
-
+
Samples 5 and 6 were from the same “pool” of cells. Looking at the
documentation for the dataset under ?WTChimeraData we see
@@ -3822,7 +3624,7 @@
Show me the solution
first TSNE plot. If you were developing single-cell library preparation
protocols you might want to preserve this effect to understand how
variation in pools leads to variation in expression, but for now, given
-that we’re investigate other effects, we’ll want to remove this as
+that we’re investigating other effects, we’ll want to remove this as
undesired technical variation.
@@ -3831,29 +3633,30 @@ Show me the solution
Correcting batch effects
-We correct the effect of samples by aid of the
+
We “correct” the effect of samples with the
correctExperiment function in the batchelor
-package and using the sample colData column as
-batch.
+package and using the sample column as batch.
R
-library(batchelor)
-set.seed(10102)
-merged <- correctExperiments(sce,
- batch=sce$sample,
- subset.row=chosen.hvgs,
- PARAM=FastMnnParam(
- merge.order=list(
+set.seed(10102)
+
+merged <- correctExperiments(
+ sce,
+ batch = sce$sample,
+ subset.row = chosen.hvgs,
+ PARAM = FastMnnParam(
+ merge.order = list(
list(1,3,5), # WT (3 replicates)
list(2,4,6) # td-Tomato (3 replicates)
)
)
)
-merged <- runTSNE(merged, dimred="corrected")
-plotTSNE(merged, colour_by="batch")
+merged <- runTSNE(merged, dimred = "corrected")
+
+plotTSNE(merged, colour_by = "batch")
Once we removed the sample batch effect, we can proceed with the
Differential Expression Analysis.
@@ -3875,7 +3678,7 @@ Challenge
-
+
It’s important to have multiple samples within each experimental
group because it helps the batch effect correction algorithm distinguish
@@ -3898,7 +3701,7 @@
Show me the solution
Differential Expression
-In order to perform a Differential Expression Analysis, we need to
+
In order to perform a differential expression analysis, we need to
identify groups of cells across samples/conditions (depending on the
experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell
@@ -3926,9 +3729,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output
# corresponds to one unique combination of these two factors.
-library(scuttle)
-summed <- aggregateAcrossCells(merged,
- id=colData(merged)[,c("celltype.mapped", "sample")])
+
+summed <- aggregateAcrossCells(
+ merged,
+ id = colData(merged)[,c("celltype.mapped", "sample")]
+)
+
summed
R
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
y
@@ -4018,7 +3824,9 @@ R
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)
@@ -4034,8 +3842,10 @@ OUTPUT<
R
-keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)
@@ -4055,6 +3865,7 @@ R
y <- calcNormFactors(y)
+
y$samples
@@ -4095,7 +3906,7 @@ Rpar(mfrow = c(2,3))
for (i in seq_len(ncol(y))) {
- plotMD(y, column=i)
+ plotMD(y, column = i)
}
RR
-limma::plotMDS(cpm(y, log=TRUE),
- col=ifelse(y$samples$tomato, "red", "blue"))
+limma::plotMDS(cpm(y, log = TRUE),
+ col = ifelse(y$samples$tomato, "red", "blue"))
We then construct a design matrix by including both the pool and the
tomato as factors. This design indicates which samples belong to which
@@ -4153,6 +3964,7 @@
R
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)
@@ -4172,7 +3984,7 @@ RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized
-linear model for each gene. The robust=TRUE parameter
+linear model for each gene. The robust = TRUE parameter
avoids distortions from highly variable clusters. The QL method includes
an additional dispersion parameter, useful to handle the uncertainty and
variability of the per-gene variance, which is not well estimated by the
@@ -4182,7 +3994,8 @@
RR
-fit <- glmQLFit(y, design, robust=TRUE)
+fit <- glmQLFit(y, design, robust = TRUE)
+
summary(fit$var.prior)
@@ -4221,7 +4034,8 @@ RR
-res <- glmQLFTest(fit, coef=ncol(design))
+res <- glmQLFTest(fit, coef = ncol(design))
+
summary(decideTests(res))
@@ -4261,14 +4075,14 @@ OUTPUT<
R
-library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)
The returned object is a list of DataFrames each with
@@ -4280,6 +4094,7 @@
R
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]
Challenge
-
+
“logFC” stands for log fold-change. Rather than reporting e.g. a
5-fold increase, it’s better to report a logFC of log(5) = 1.61.
@@ -4336,26 +4151,29 @@
Show me the solution
With DA we look for differences in cluster abundance across
conditions (the tomato injection in our case), rather than differences
in gene expression.
-We first setup some code and variables for further analysis, like
-quantifying the number of cells per each cell type and fit a model to
-catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this
-time we start our analysis on the computed abundances and without
+
Our first steps are quantifying the number of cells per each cell
+type and fitting a model to catch differences between the injected cells
+and the background.
+The process is very similar differential expression modeling, but
+this time we start our analysis on the computed abundances and without
normalizing the data with TMM.
R
-library(edgeR)
-abundances <- table(merged$celltype.mapped, merged$sample)
+abundances <- table(merged$celltype.mapped, merged$sample)
+
abundances <- unclass(abundances)
-# Attaching some column metadata.
+
extra.info <- colData(merged)[match(colnames(abundances), merged$sample),]
-y.ab <- DGEList(abundances, samples=extra.info)
+
+y.ab <- DGEList(abundances, samples = extra.info)
design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples)
-y.ab <- estimateDisp(y.ab, design, trend="none")
-fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)
+
+y.ab <- estimateDisp(y.ab, design, trend = "none")
+
+fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)
Background on compositional effect
@@ -4401,6 +4219,7 @@ R
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
@@ -4413,11 +4232,11 @@ OUTPUT<
R
-y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))
@@ -4433,7 +4252,7 @@ OUTPUT<
R
-topTags(res2, n=10)
+topTags(res2, n = 10)
OUTPUT
@@ -4469,7 +4288,8 @@ Testing against a log-fold
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
@@ -4550,70 +4370,72 @@ OUTPUT<
[8] base
other attached packages:
- [1] edgeR_4.2.0 limma_3.60.2
- [3] batchelor_1.20.0 scran_1.32.0
- [5] scater_1.32.0 ggplot2_3.5.1
- [7] scuttle_1.14.0 MouseGastrulationData_1.18.0
- [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0
-[11] SummarizedExperiment_1.34.0 Biobase_2.64.0
-[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1
-[15] IRanges_2.38.0 S4Vectors_0.42.0
-[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0
-[19] matrixStats_1.3.0 BiocStyle_2.32.0
+ [1] pheatmap_1.0.12 scran_1.32.0
+ [3] scater_1.32.0 ggplot2_3.5.1
+ [5] scuttle_1.14.0 edgeR_4.2.0
+ [7] limma_3.60.2 batchelor_1.20.0
+ [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0
+[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0
+[13] Biobase_2.64.0 GenomicRanges_1.56.0
+[15] GenomeInfoDb_1.40.1 IRanges_2.38.0
+[17] S4Vectors_0.42.0 BiocGenerics_0.50.0
+[19] MatrixGenerics_1.16.0 matrixStats_1.3.0
+[21] BiocStyle_2.32.0
loaded via a namespace (and not attached):
- [1] DBI_1.2.3 formatR_1.14
- [3] gridExtra_2.3 rlang_1.1.3
- [5] magrittr_2.0.3 compiler_4.4.1
- [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0
- [9] png_0.1-8 vctrs_0.6.5
- [11] pkgconfig_2.0.3 crayon_1.5.2
- [13] fastmap_1.2.0 dbplyr_2.5.0
- [15] magick_2.8.3 XVector_0.44.0
- [17] labeling_0.4.3 utf8_1.2.4
- [19] rmarkdown_2.27 UCSC.utils_1.0.0
- [21] ggbeeswarm_0.7.2 purrr_1.0.2
- [23] bit_4.0.5 bluster_1.14.0
- [25] xfun_0.44 zlibbioc_1.50.0
- [27] cachem_1.1.0 beachmat_2.20.0
- [29] jsonlite_1.8.8 blob_1.2.4
- [31] highr_0.11 DelayedArray_0.30.1
- [33] BiocParallel_1.38.0 cluster_2.1.6
- [35] irlba_2.3.5.1 parallel_4.4.1
- [37] R6_2.5.1 Rcpp_1.0.12
- [39] knitr_1.47 splines_4.4.1
- [41] igraph_2.0.3 Matrix_1.7-0
- [43] tidyselect_1.2.1 viridis_0.6.5
- [45] abind_1.4-5 yaml_2.3.8
- [47] codetools_0.2-20 curl_5.2.1
- [49] lattice_0.22-6 tibble_3.2.1
- [51] withr_3.0.0 KEGGREST_1.44.0
- [53] BumpyMatrix_1.12.0 Rtsne_0.17
- [55] evaluate_0.23 BiocFileCache_2.12.0
- [57] ExperimentHub_2.12.0 Biostrings_2.72.1
- [59] pillar_1.9.0 BiocManager_1.30.23
- [61] filelock_1.0.3 renv_1.0.7
- [63] generics_0.1.3 BiocVersion_3.19.1
- [65] sparseMatrixStats_1.16.0 munsell_0.5.1
- [67] scales_1.3.0 glue_1.7.0
- [69] metapod_1.12.0 tools_4.4.1
- [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0
- [73] ScaledMatrix_1.12.0 locfit_1.5-9.9
- [75] cowplot_1.1.3 grid_4.4.1
- [77] AnnotationDbi_1.66.0 colorspace_2.1-0
- [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0
- [81] BiocSingular_1.20.0 vipor_0.4.7
- [83] cli_3.6.2 rsvd_1.0.5
- [85] rappdirs_0.3.3 fansi_1.0.6
- [87] viridisLite_0.4.2 S4Arrays_1.4.1
- [89] dplyr_1.1.4 ResidualMatrix_1.14.0
- [91] gtable_0.3.5 digest_0.6.35
- [93] dqrng_0.4.1 SparseArray_1.4.8
- [95] ggrepel_0.9.5 farver_2.1.2
- [97] rjson_0.2.21 memoise_2.0.1
- [99] htmltools_0.5.8.1 lifecycle_1.0.4
-[101] httr_1.4.7 statmod_1.5.0
-[103] mime_0.12 bit64_4.0.5
+ [1] RColorBrewer_1.1-3 jsonlite_1.8.8
+ [3] magrittr_2.0.3 ggbeeswarm_0.7.2
+ [5] magick_2.8.3 farver_2.1.2
+ [7] rmarkdown_2.27 zlibbioc_1.50.0
+ [9] vctrs_0.6.5 memoise_2.0.1
+ [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1
+ [13] S4Arrays_1.4.1 AnnotationHub_3.12.0
+ [15] curl_5.2.1 BiocNeighbors_1.22.0
+ [17] SparseArray_1.4.8 cachem_1.1.0
+ [19] ResidualMatrix_1.14.0 igraph_2.0.3
+ [21] mime_0.12 lifecycle_1.0.4
+ [23] pkgconfig_2.0.3 rsvd_1.0.5
+ [25] Matrix_1.7-0 R6_2.5.1
+ [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12
+ [29] digest_0.6.35 colorspace_2.1-0
+ [31] AnnotationDbi_1.66.0 dqrng_0.4.1
+ [33] irlba_2.3.5.1 ExperimentHub_2.12.0
+ [35] RSQLite_2.3.7 beachmat_2.20.0
+ [37] filelock_1.0.3 labeling_0.4.3
+ [39] fansi_1.0.6 httr_1.4.7
+ [41] abind_1.4-5 compiler_4.4.1
+ [43] bit64_4.0.5 withr_3.0.0
+ [45] BiocParallel_1.38.0 viridis_0.6.5
+ [47] DBI_1.2.3 highr_0.11
+ [49] rappdirs_0.3.3 DelayedArray_0.30.1
+ [51] rjson_0.2.21 bluster_1.14.0
+ [53] tools_4.4.1 vipor_0.4.7
+ [55] beeswarm_0.4.0 glue_1.7.0
+ [57] grid_4.4.1 Rtsne_0.17
+ [59] cluster_2.1.6 generics_0.1.3
+ [61] gtable_0.3.5 BiocSingular_1.20.0
+ [63] ScaledMatrix_1.12.0 metapod_1.12.0
+ [65] utf8_1.2.4 XVector_0.44.0
+ [67] ggrepel_0.9.5 BiocVersion_3.19.1
+ [69] pillar_1.9.0 BumpyMatrix_1.12.0
+ [71] splines_4.4.1 dplyr_1.1.4
+ [73] BiocFileCache_2.12.0 lattice_0.22-6
+ [75] renv_1.0.7 bit_4.0.5
+ [77] tidyselect_1.2.1 locfit_1.5-9.9
+ [79] Biostrings_2.72.1 knitr_1.47
+ [81] gridExtra_2.3 xfun_0.44
+ [83] statmod_1.5.0 UCSC.utils_1.0.0
+ [85] yaml_2.3.8 evaluate_0.23
+ [87] codetools_0.2-20 tibble_3.2.1
+ [89] BiocManager_1.30.23 cli_3.6.2
+ [91] munsell_0.5.1 Rcpp_1.0.12
+ [93] dbplyr_2.5.0 png_0.1-8
+ [95] parallel_4.4.1 blob_1.2.4
+ [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2
+ [99] scales_1.3.0 purrr_1.0.2
+[101] crayon_1.5.2 rlang_1.1.3
+[103] cowplot_1.1.3 KEGGREST_1.44.0
+[105] formatR_1.14
Exercises
@@ -4636,7 +4458,7 @@ Exercise 1: Replicates
-
+
Remember, you can subset SingleCellExperiments with logical indices,
just like a matrix. You can also access their column data with the
@@ -4650,7 +4472,7 @@
Give me a hint
-
+
R
@@ -4658,11 +4480,12 @@ R
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
@@ -4686,7 +4509,7 @@ Exercise 2:
-
+
You can just hand pheatmap() a matrix as its only
argument. It has a million options, but the defaults are usually pretty
@@ -4700,15 +4523,13 @@
Give me a hint
-
+
R
-library(pheatmap)
-
-pheatmap(y.ab$counts)
+pheatmap(y.ab$counts)
The top DA result was a decrease in ExE ectoderm in the tomato
condition, which you can sort of see, especially if you
@@ -6470,7 +6291,7 @@
Exercise 1
-
+
R
@@ -6503,7 +6324,7 @@ Exercise 2
-
+
R
@@ -6535,7 +6356,7 @@ Exercise 3
-
+
R
diff --git a/index.html b/index.html
index 24dd0f3..6dfcdaf 100644
--- a/index.html
+++ b/index.html
@@ -372,7 +372,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -398,7 +398,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN
website.
@@ -429,7 +429,7 @@
If you already have R and RStudio installed
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -453,7 +453,7 @@
If you don’t have R and RStudio installed
-
+
- Download R from the CRAN website.
- Select the
.pkg file for the latest R version
@@ -486,7 +486,7 @@
Install R using your package manager and RStudio
-
+
- Follow the instructions for your distribution from CRAN, they provide
information to get the most recent version of R for common
diff --git a/instructor/aio.html b/instructor/aio.html
index 4714d2d..f7a8835 100644
--- a/instructor/aio.html
+++ b/instructor/aio.html
@@ -1040,7 +1040,7 @@
Challenge
-
+
R
@@ -1120,7 +1120,7 @@ OUTPUT<
-
+
Whenever your code involves the generation of random numbers, it’s a
good practice to set the random seed in R with
@@ -1353,7 +1353,7 @@
Challenge
-
+
R
@@ -1629,7 +1629,7 @@ Challenge
-
+
Spike-ins are deliberately-introduced exogeneous RNA from an exotic
or synthetic source at a known concentration. This provides a known
@@ -1924,7 +1924,7 @@
Challenge
-
+
Mathematically, this would require the data to fall on a
two-dimensional plane (for linear methods like PCA) or a smooth 2D
@@ -2094,7 +2094,7 @@
Exercise 2: Dimensionality Reduction
-
+
First subset the object to include only highly variable genes
(sce2 <- sce[hvg.sce.var,]) and then apply the
@@ -2380,7 +2380,7 @@
Challenge
-
+
We see in the help documentation for ?clusterCells that
all of the clustering algorithm details are handled through the
@@ -2533,7 +2533,7 @@
Challenge
-
+
One important reason why is because averages over all other clusters
can be sensitive to the cell type composition. If a rare cell type shows
@@ -2826,7 +2826,7 @@
Challenge
-
+
Use BiocParallel and the BPPARAM argument!
This example will set it to use four cores on your laptop, but you can
@@ -3015,7 +3015,9 @@
OUTPUT<
a more general diagnostic check is to compare the assigned labels to
cluster identities, under the expectation that most cells of a single
cluster would have the same label (or, if multiple labels are present,
-they should at least represent closely related cell states).
+they should at least represent closely related cell states). We only
+print out the top-left corner of the table here, but you should try
+looking at the whole thing:
R
@@ -3024,231 +3026,23 @@ R
tab <- table(new.labels, sce$celltype.mapped)
-tab
+tab[1:4,1:4]
OUTPUT
-new.labels Allantois Blood progenitors 1
- Allantois 28 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 1 0
- Gut 0 0
- Haematoendothelial progenitors 1 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 11 4
- NMP 0 0
- Paraxial mesoderm 5 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Blood progenitors 2 Cardiomyocytes
- Allantois 0 0
- Cardiomyocytes 0 27
- Endothelium 0 0
- Erythroid2 1 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 0 0
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 6 1
- NMP 0 0
- Paraxial mesoderm 0 4
- Pharyngeal mesoderm 0 6
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Caudal epiblast Caudal Mesoderm Def. endoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 1 0 2
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 0
- Neural crest 0 0 1
- NMP 0 0 0
- Paraxial mesoderm 0 0 1
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 1 1
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Doublet Endothelium Erythroid1 Erythroid2
- Allantois 0 0 0 0
- Cardiomyocytes 0 0 0 0
- Endothelium 0 13 0 0
- Erythroid2 0 0 14 21
- Erythroid3 0 0 6 5
- ExE mesoderm 0 0 0 0
- Forebrain/Midbrain/Hindbrain 6 0 0 0
- Gut 0 0 0 0
- Haematoendothelial progenitors 0 19 0 0
- Intermediate mesoderm 13 0 0 0
- Mesenchyme 0 0 0 0
- Neural crest 20 1 3 0
- NMP 0 0 0 0
- Paraxial mesoderm 5 0 0 0
- Pharyngeal mesoderm 1 0 0 0
- Somitic mesoderm 1 0 0 0
- Spinal cord 1 0 0 0
- Surface ectoderm 0 0 0 0
-
-new.labels Erythroid3 ExE mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 51 0
- Erythroid3 49 0
- ExE mesoderm 0 23
- Forebrain/Midbrain/Hindbrain 0 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 16
- Mesenchyme 0 0
- Neural crest 0 17
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 0
+new.labels Allantois Blood progenitors 1 Blood progenitors 2
+ Allantois 28 0 0
+ Cardiomyocytes 0 0 0
+ Endothelium 0 0 0
+ Erythroid2 0 0 1
-new.labels Forebrain/Midbrain/Hindbrain Gut
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 83 4
- Gut 0 5
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 0
- Mesenchyme 0 0
- Neural crest 20 0
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 0
- Spinal cord 0 0
- Surface ectoderm 0 12
-
-new.labels Haematoendothelial progenitors
- Allantois 0
- Cardiomyocytes 0
- Endothelium 0
- Erythroid2 0
- Erythroid3 0
- ExE mesoderm 0
- Forebrain/Midbrain/Hindbrain 0
- Gut 0
- Haematoendothelial progenitors 16
- Intermediate mesoderm 0
- Mesenchyme 0
- Neural crest 8
- NMP 0
- Paraxial mesoderm 2
- Pharyngeal mesoderm 0
- Somitic mesoderm 0
- Spinal cord 0
- Surface ectoderm 0
-
-new.labels Intermediate mesoderm Mesenchyme Neural crest
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 1 7 0
- Forebrain/Midbrain/Hindbrain 0 3 0
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 9 15 0
- Mesenchyme 0 69 0
- Neural crest 4 4 10
- NMP 0 0 0
- Paraxial mesoderm 0 7 0
- Pharyngeal mesoderm 0 13 0
- Somitic mesoderm 2 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels NMP Paraxial mesoderm Pharyngeal mesoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 0 0
- Erythroid2 0 0 0
- Erythroid3 0 0 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 41 2 4
- Gut 0 0 0
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 1 5
- Mesenchyme 0 0 0
- Neural crest 5 20 10
- NMP 16 0 0
- Paraxial mesoderm 0 43 3
- Pharyngeal mesoderm 0 0 18
- Somitic mesoderm 0 0 0
- Spinal cord 0 0 0
- Surface ectoderm 0 0 0
-
-new.labels Rostral neurectoderm Somitic mesoderm
- Allantois 0 0
- Cardiomyocytes 0 0
- Endothelium 0 0
- Erythroid2 0 0
- Erythroid3 0 0
- ExE mesoderm 0 0
- Forebrain/Midbrain/Hindbrain 11 1
- Gut 0 0
- Haematoendothelial progenitors 0 0
- Intermediate mesoderm 0 2
- Mesenchyme 0 0
- Neural crest 0 7
- NMP 0 0
- Paraxial mesoderm 0 0
- Pharyngeal mesoderm 0 0
- Somitic mesoderm 0 25
- Spinal cord 0 0
- Surface ectoderm 0 0
-
-new.labels Spinal cord Stripped Surface ectoderm
- Allantois 0 0 0
- Cardiomyocytes 0 0 0
- Endothelium 0 1 0
- Erythroid2 0 1 0
- Erythroid3 0 1 0
- ExE mesoderm 0 0 0
- Forebrain/Midbrain/Hindbrain 37 0 17
- Gut 0 0 2
- Haematoendothelial progenitors 0 0 0
- Intermediate mesoderm 0 0 0
- Mesenchyme 0 0 2
- Neural crest 3 5 6
- NMP 0 0 0
- Paraxial mesoderm 0 0 0
- Pharyngeal mesoderm 0 0 0
- Somitic mesoderm 0 0 0
- Spinal cord 9 0 0
- Surface ectoderm 0 0 20
+new.labels Cardiomyocytes
+ Allantois 0
+ Cardiomyocytes 27
+ Endothelium 0
+ Erythroid2 0
As a diagnostic measure, we examine the distribution of AUCs across
cells for each label. In heterogeneous populations, the distribution for
@@ -3292,7 +3086,7 @@
Challenge
-
+
The example that jumps out most strongly to the eye is ExE endoderm,
which doesn’t show clear separate modes. Simultaneously, Endothelium
@@ -3441,7 +3235,7 @@
Exercise 1: Clustering
-
+
The NNGraphParam constructor has an argument
cluster.args. This allows to specify arguments passed on to
@@ -3458,7 +3252,7 @@
Give me a hint
-
+
TODO
@@ -3487,7 +3281,7 @@ Exercise 2: Cluster annotation
-
+
Use the goana() function from the limma
package to identify GO BP terms that are overrepresented in the list of
@@ -3501,7 +3295,7 @@
Give me a hint
-
+
TODO
@@ -3531,7 +3325,7 @@ Exercise 3: Workflow
-
+
TODO
@@ -3603,7 +3397,7 @@ Key Points
-->
Content from Multi-sample analyses
-Last updated on 2024-09-06 |
+
Last updated on 2024-09-08 |
Edit this page
Estimated time: 45 minutes
@@ -3661,13 +3455,23 @@ Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.
We start by loading the data and doing a quick exploratory analysis,
essentially applying the normalization and visualization techniques that
-we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this
+time we’re selecting samples 5 to 10, not just 5 by itself.
R
library(MouseGastrulationData)
-sce <- WTChimeraData(samples=5:10, type = "processed")
+library(batchelor)
+library(edgeR)
+library(scater)
+library(ggplot2)
+library(scran)
+library(pheatmap)
+library(scuttle)
+
+sce <- WTChimeraData(samples = 5:10, type = "processed")
+
sce
@@ -3742,15 +3546,12 @@ OUTPUT<
R
-library(scater)
-library(ggplot2)
-library(scran)
+drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
-# remove doublets
-drop <- sce$celltype.mapped %in% c("stripped", "Doublet")
sce <- sce[,!drop]
set.seed(29482)
+
idx <- unlist(tapply(colnames(sce), sce$sample, function(x) {
perc <- round(0.50 * length(x))
sample(x, perc)
@@ -3758,23 +3559,24 @@ R
sce <- sce[,idx]
-We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps,
+and visualize them in a tSNE plot.
R
-# normalization
-sce <- logNormCounts(sce)
+sce <- logNormCounts(sce)
+
+dec <- modelGeneVar(sce, block = sce$sample)
-# identify highly variable genes
-dec <- modelGeneVar(sce, block=sce$sample)
chosen.hvgs <- dec$bio > 0
-# dimensionality reduction
sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000)
+
sce <- runTSNE(sce, dimred = "PCA")
sce$sample <- as.factor(sce$sample)
+
plotTSNE(sce, colour_by = "sample")
Challenge
-
+
Samples 5 and 6 were from the same “pool” of cells. Looking at the
documentation for the dataset under ?WTChimeraData we see
@@ -3822,7 +3624,7 @@
Show me the solution
first TSNE plot. If you were developing single-cell library preparation
protocols you might want to preserve this effect to understand how
variation in pools leads to variation in expression, but for now, given
-that we’re investigate other effects, we’ll want to remove this as
+that we’re investigating other effects, we’ll want to remove this as
undesired technical variation.
@@ -3831,29 +3633,30 @@ Show me the solution
Correcting batch effects
-We correct the effect of samples by aid of the
+
We “correct” the effect of samples with the
correctExperiment function in the batchelor
-package and using the sample colData column as
-batch.
+package and using the sample column as batch.
R
-library(batchelor)
-set.seed(10102)
-merged <- correctExperiments(sce,
- batch=sce$sample,
- subset.row=chosen.hvgs,
- PARAM=FastMnnParam(
- merge.order=list(
+set.seed(10102)
+
+merged <- correctExperiments(
+ sce,
+ batch = sce$sample,
+ subset.row = chosen.hvgs,
+ PARAM = FastMnnParam(
+ merge.order = list(
list(1,3,5), # WT (3 replicates)
list(2,4,6) # td-Tomato (3 replicates)
)
)
)
-merged <- runTSNE(merged, dimred="corrected")
-plotTSNE(merged, colour_by="batch")
+merged <- runTSNE(merged, dimred = "corrected")
+
+plotTSNE(merged, colour_by = "batch")
Once we removed the sample batch effect, we can proceed with the
Differential Expression Analysis.
@@ -3875,7 +3678,7 @@ Challenge
-
+
It’s important to have multiple samples within each experimental
group because it helps the batch effect correction algorithm distinguish
@@ -3898,7 +3701,7 @@
Show me the solution
Differential Expression
-In order to perform a Differential Expression Analysis, we need to
+
In order to perform a differential expression analysis, we need to
identify groups of cells across samples/conditions (depending on the
experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell
@@ -3926,9 +3729,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output
# corresponds to one unique combination of these two factors.
-library(scuttle)
-summed <- aggregateAcrossCells(merged,
- id=colData(merged)[,c("celltype.mapped", "sample")])
+
+summed <- aggregateAcrossCells(
+ merged,
+ id = colData(merged)[,c("celltype.mapped", "sample")]
+)
+
summed
R
label <- "Mesenchyme"
-current <- summed[,label==summed$celltype.mapped]
-# Creating up a DGEList object for use in edgeR:
-library(edgeR)
-y <- DGEList(counts(current), samples=colData(current))
+current <- summed[,label == summed$celltype.mapped]
+
+y <- DGEList(counts(current), samples = colData(current))
+
y
@@ -4018,7 +3824,9 @@ R
discarded <- current$ncells < 10
+
y <- y[,!discarded]
+
summary(discarded)
@@ -4034,8 +3842,10 @@ OUTPUT<
R
-keep <- filterByExpr(y, group=current$tomato)
+keep <- filterByExpr(y, group = current$tomato)
+
y <- y[keep,]
+
summary(keep)
@@ -4055,6 +3865,7 @@ R
y <- calcNormFactors(y)
+
y$samples
@@ -4095,7 +3906,7 @@ Rpar(mfrow = c(2,3))
for (i in seq_len(ncol(y))) {
- plotMD(y, column=i)
+ plotMD(y, column = i)
}
RR
-limma::plotMDS(cpm(y, log=TRUE),
- col=ifelse(y$samples$tomato, "red", "blue"))
+limma::plotMDS(cpm(y, log = TRUE),
+ col = ifelse(y$samples$tomato, "red", "blue"))
We then construct a design matrix by including both the pool and the
tomato as factors. This design indicates which samples belong to which
@@ -4153,6 +3964,7 @@
R
y <- estimateDisp(y, design)
+
summary(y$trended.dispersion)
@@ -4172,7 +3984,7 @@ RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized
-linear model for each gene. The robust=TRUE parameter
+linear model for each gene. The robust = TRUE parameter
avoids distortions from highly variable clusters. The QL method includes
an additional dispersion parameter, useful to handle the uncertainty and
variability of the per-gene variance, which is not well estimated by the
@@ -4182,7 +3994,8 @@
RR
-fit <- glmQLFit(y, design, robust=TRUE)
+fit <- glmQLFit(y, design, robust = TRUE)
+
summary(fit$var.prior)
@@ -4221,7 +4034,8 @@ RR
-res <- glmQLFTest(fit, coef=ncol(design))
+res <- glmQLFTest(fit, coef = ncol(design))
+
summary(decideTests(res))
@@ -4261,14 +4075,14 @@ OUTPUT<
R
-library(scran)
-summed.filt <- summed[,summed$ncells >= 10]
+summed.filt <- summed[,summed$ncells >= 10]
-de.results <- pseudoBulkDGE(summed.filt,
- label=summed.filt$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt,
+ label = summed.filt$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt$tomato
)
The returned object is a list of DataFrames each with
@@ -4280,6 +4094,7 @@
R
cur.results <- de.results[["Allantois"]]
+
cur.results[order(cur.results$PValue),]
Challenge
-
+
“logFC” stands for log fold-change. Rather than reporting e.g. a
5-fold increase, it’s better to report a logFC of log(5) = 1.61.
@@ -4336,26 +4151,29 @@
Show me the solution
With DA we look for differences in cluster abundance across
conditions (the tomato injection in our case), rather than differences
in gene expression.
-We first setup some code and variables for further analysis, like
-quantifying the number of cells per each cell type and fit a model to
-catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this
-time we start our analysis on the computed abundances and without
+
Our first steps are quantifying the number of cells per each cell
+type and fitting a model to catch differences between the injected cells
+and the background.
+The process is very similar differential expression modeling, but
+this time we start our analysis on the computed abundances and without
normalizing the data with TMM.
R
-library(edgeR)
-abundances <- table(merged$celltype.mapped, merged$sample)
+abundances <- table(merged$celltype.mapped, merged$sample)
+
abundances <- unclass(abundances)
-# Attaching some column metadata.
+
extra.info <- colData(merged)[match(colnames(abundances), merged$sample),]
-y.ab <- DGEList(abundances, samples=extra.info)
+
+y.ab <- DGEList(abundances, samples = extra.info)
design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples)
-y.ab <- estimateDisp(y.ab, design, trend="none")
-fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)
+
+y.ab <- estimateDisp(y.ab, design, trend = "none")
+
+fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)
Background on compositional effect
@@ -4401,6 +4219,7 @@ R
y.ab2 <- calcNormFactors(y.ab)
+
y.ab2$samples$norm.factors
@@ -4413,11 +4232,11 @@ OUTPUT<
R
-y.ab2 <- estimateDisp(y.ab2, design, trend="none")
+y.ab2 <- estimateDisp(y.ab2, design, trend = "none")
-fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE)
+fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE)
-res2 <- glmQLFTest(fit.ab2, coef=ncol(design))
+res2 <- glmQLFTest(fit.ab2, coef = ncol(design))
summary(decideTests(res2))
@@ -4433,7 +4252,7 @@ OUTPUT<
R
-topTags(res2, n=10)
+topTags(res2, n = 10)
OUTPUT
@@ -4469,7 +4288,8 @@ Testing against a log-fold
R
-res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1)
+res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1)
+
summary(decideTests(res.lfc))
@@ -4550,70 +4370,72 @@ OUTPUT<
[8] base
other attached packages:
- [1] edgeR_4.2.0 limma_3.60.2
- [3] batchelor_1.20.0 scran_1.32.0
- [5] scater_1.32.0 ggplot2_3.5.1
- [7] scuttle_1.14.0 MouseGastrulationData_1.18.0
- [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0
-[11] SummarizedExperiment_1.34.0 Biobase_2.64.0
-[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1
-[15] IRanges_2.38.0 S4Vectors_0.42.0
-[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0
-[19] matrixStats_1.3.0 BiocStyle_2.32.0
+ [1] pheatmap_1.0.12 scran_1.32.0
+ [3] scater_1.32.0 ggplot2_3.5.1
+ [5] scuttle_1.14.0 edgeR_4.2.0
+ [7] limma_3.60.2 batchelor_1.20.0
+ [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0
+[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0
+[13] Biobase_2.64.0 GenomicRanges_1.56.0
+[15] GenomeInfoDb_1.40.1 IRanges_2.38.0
+[17] S4Vectors_0.42.0 BiocGenerics_0.50.0
+[19] MatrixGenerics_1.16.0 matrixStats_1.3.0
+[21] BiocStyle_2.32.0
loaded via a namespace (and not attached):
- [1] DBI_1.2.3 formatR_1.14
- [3] gridExtra_2.3 rlang_1.1.3
- [5] magrittr_2.0.3 compiler_4.4.1
- [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0
- [9] png_0.1-8 vctrs_0.6.5
- [11] pkgconfig_2.0.3 crayon_1.5.2
- [13] fastmap_1.2.0 dbplyr_2.5.0
- [15] magick_2.8.3 XVector_0.44.0
- [17] labeling_0.4.3 utf8_1.2.4
- [19] rmarkdown_2.27 UCSC.utils_1.0.0
- [21] ggbeeswarm_0.7.2 purrr_1.0.2
- [23] bit_4.0.5 bluster_1.14.0
- [25] xfun_0.44 zlibbioc_1.50.0
- [27] cachem_1.1.0 beachmat_2.20.0
- [29] jsonlite_1.8.8 blob_1.2.4
- [31] highr_0.11 DelayedArray_0.30.1
- [33] BiocParallel_1.38.0 cluster_2.1.6
- [35] irlba_2.3.5.1 parallel_4.4.1
- [37] R6_2.5.1 Rcpp_1.0.12
- [39] knitr_1.47 splines_4.4.1
- [41] igraph_2.0.3 Matrix_1.7-0
- [43] tidyselect_1.2.1 viridis_0.6.5
- [45] abind_1.4-5 yaml_2.3.8
- [47] codetools_0.2-20 curl_5.2.1
- [49] lattice_0.22-6 tibble_3.2.1
- [51] withr_3.0.0 KEGGREST_1.44.0
- [53] BumpyMatrix_1.12.0 Rtsne_0.17
- [55] evaluate_0.23 BiocFileCache_2.12.0
- [57] ExperimentHub_2.12.0 Biostrings_2.72.1
- [59] pillar_1.9.0 BiocManager_1.30.23
- [61] filelock_1.0.3 renv_1.0.7
- [63] generics_0.1.3 BiocVersion_3.19.1
- [65] sparseMatrixStats_1.16.0 munsell_0.5.1
- [67] scales_1.3.0 glue_1.7.0
- [69] metapod_1.12.0 tools_4.4.1
- [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0
- [73] ScaledMatrix_1.12.0 locfit_1.5-9.9
- [75] cowplot_1.1.3 grid_4.4.1
- [77] AnnotationDbi_1.66.0 colorspace_2.1-0
- [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0
- [81] BiocSingular_1.20.0 vipor_0.4.7
- [83] cli_3.6.2 rsvd_1.0.5
- [85] rappdirs_0.3.3 fansi_1.0.6
- [87] viridisLite_0.4.2 S4Arrays_1.4.1
- [89] dplyr_1.1.4 ResidualMatrix_1.14.0
- [91] gtable_0.3.5 digest_0.6.35
- [93] dqrng_0.4.1 SparseArray_1.4.8
- [95] ggrepel_0.9.5 farver_2.1.2
- [97] rjson_0.2.21 memoise_2.0.1
- [99] htmltools_0.5.8.1 lifecycle_1.0.4
-[101] httr_1.4.7 statmod_1.5.0
-[103] mime_0.12 bit64_4.0.5
+ [1] RColorBrewer_1.1-3 jsonlite_1.8.8
+ [3] magrittr_2.0.3 ggbeeswarm_0.7.2
+ [5] magick_2.8.3 farver_2.1.2
+ [7] rmarkdown_2.27 zlibbioc_1.50.0
+ [9] vctrs_0.6.5 memoise_2.0.1
+ [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1
+ [13] S4Arrays_1.4.1 AnnotationHub_3.12.0
+ [15] curl_5.2.1 BiocNeighbors_1.22.0
+ [17] SparseArray_1.4.8 cachem_1.1.0
+ [19] ResidualMatrix_1.14.0 igraph_2.0.3
+ [21] mime_0.12 lifecycle_1.0.4
+ [23] pkgconfig_2.0.3 rsvd_1.0.5
+ [25] Matrix_1.7-0 R6_2.5.1
+ [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12
+ [29] digest_0.6.35 colorspace_2.1-0
+ [31] AnnotationDbi_1.66.0 dqrng_0.4.1
+ [33] irlba_2.3.5.1 ExperimentHub_2.12.0
+ [35] RSQLite_2.3.7 beachmat_2.20.0
+ [37] filelock_1.0.3 labeling_0.4.3
+ [39] fansi_1.0.6 httr_1.4.7
+ [41] abind_1.4-5 compiler_4.4.1
+ [43] bit64_4.0.5 withr_3.0.0
+ [45] BiocParallel_1.38.0 viridis_0.6.5
+ [47] DBI_1.2.3 highr_0.11
+ [49] rappdirs_0.3.3 DelayedArray_0.30.1
+ [51] rjson_0.2.21 bluster_1.14.0
+ [53] tools_4.4.1 vipor_0.4.7
+ [55] beeswarm_0.4.0 glue_1.7.0
+ [57] grid_4.4.1 Rtsne_0.17
+ [59] cluster_2.1.6 generics_0.1.3
+ [61] gtable_0.3.5 BiocSingular_1.20.0
+ [63] ScaledMatrix_1.12.0 metapod_1.12.0
+ [65] utf8_1.2.4 XVector_0.44.0
+ [67] ggrepel_0.9.5 BiocVersion_3.19.1
+ [69] pillar_1.9.0 BumpyMatrix_1.12.0
+ [71] splines_4.4.1 dplyr_1.1.4
+ [73] BiocFileCache_2.12.0 lattice_0.22-6
+ [75] renv_1.0.7 bit_4.0.5
+ [77] tidyselect_1.2.1 locfit_1.5-9.9
+ [79] Biostrings_2.72.1 knitr_1.47
+ [81] gridExtra_2.3 xfun_0.44
+ [83] statmod_1.5.0 UCSC.utils_1.0.0
+ [85] yaml_2.3.8 evaluate_0.23
+ [87] codetools_0.2-20 tibble_3.2.1
+ [89] BiocManager_1.30.23 cli_3.6.2
+ [91] munsell_0.5.1 Rcpp_1.0.12
+ [93] dbplyr_2.5.0 png_0.1-8
+ [95] parallel_4.4.1 blob_1.2.4
+ [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2
+ [99] scales_1.3.0 purrr_1.0.2
+[101] crayon_1.5.2 rlang_1.1.3
+[103] cowplot_1.1.3 KEGGREST_1.44.0
+[105] formatR_1.14
Exercises
@@ -4636,7 +4458,7 @@ Exercise 1: Replicates
-
+
Remember, you can subset SingleCellExperiments with logical indices,
just like a matrix. You can also access their column data with the
@@ -4650,7 +4472,7 @@
Give me a hint
-
+
R
@@ -4658,11 +4480,12 @@ R
summed.filt.subset = summed.filt[,summed.filt$pool != 3]
-de.results <- pseudoBulkDGE(summed.filt.subset,
- label=summed.filt.subset$celltype.mapped,
- design=~factor(pool) + tomato,
- coef="tomatoTRUE",
- condition=summed.filt.subset$tomato
+de.results <- pseudoBulkDGE(
+ summed.filt.subset,
+ label = summed.filt.subset$celltype.mapped,
+ design = ~factor(pool) + tomato,
+ coef = "tomatoTRUE",
+ condition = summed.filt.subset$tomato
)
@@ -4686,7 +4509,7 @@ Exercise 2:
-
+
You can just hand pheatmap() a matrix as its only
argument. It has a million options, but the defaults are usually pretty
@@ -4700,15 +4523,13 @@
Give me a hint
-
+
R
-library(pheatmap)
-
-pheatmap(y.ab$counts)
+pheatmap(y.ab$counts)
The top DA result was a decrease in ExE ectoderm in the tomato
condition, which you can sort of see, especially if you
@@ -6470,7 +6291,7 @@
Exercise 1
-
+
R
@@ -6503,7 +6324,7 @@ Exercise 2
-
+
R
@@ -6535,7 +6356,7 @@ Exercise 3
-
+
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -398,7 +398,7 @@
If you don’t have R and RStudio installed
-+- Download R from the CRAN website. @@ -429,7 +429,7 @@
- Open RStudio, and click on “Help” > “Check for updates”. If a new
version is available, quit RStudio, and download the latest version for
@@ -453,7 +453,7 @@
If you don’t have R and RStudio installed
-+- Download R from the CRAN website.
- Select the
.pkgfile for the latest R version
@@ -486,7 +486,7 @@ - Follow the instructions for your distribution from CRAN, they provide
information to get the most recent version of R for common
diff --git a/instructor/aio.html b/instructor/aio.html
index 4714d2d..f7a8835 100644
--- a/instructor/aio.html
+++ b/instructor/aio.html
@@ -1040,7 +1040,7 @@
Challenge
-+R @@ -1120,7 +1120,7 @@
OUTPUT< -
+Whenever your code involves the generation of random numbers, it’s a good practice to set the random seed in R with @@ -1353,7 +1353,7 @@
Challenge
-+R @@ -1629,7 +1629,7 @@
Challenge
-+Spike-ins are deliberately-introduced exogeneous RNA from an exotic or synthetic source at a known concentration. This provides a known @@ -1924,7 +1924,7 @@
Challenge
-+Mathematically, this would require the data to fall on a two-dimensional plane (for linear methods like PCA) or a smooth 2D @@ -2094,7 +2094,7 @@
Exercise 2: Dimensionality Reduction
-+First subset the object to include only highly variable genes (
sce2 <- sce[hvg.sce.var,]) and then apply the @@ -2380,7 +2380,7 @@Challenge
-+We see in the help documentation for
?clusterCellsthat all of the clustering algorithm details are handled through the @@ -2533,7 +2533,7 @@Challenge
-+One important reason why is because averages over all other clusters can be sensitive to the cell type composition. If a rare cell type shows @@ -2826,7 +2826,7 @@
Challenge
-+Use
BiocParalleland theBPPARAMargument! This example will set it to use four cores on your laptop, but you can @@ -3015,7 +3015,9 @@OUTPUT< a more general diagnostic check is to compare the assigned labels to cluster identities, under the expectation that most cells of a single cluster would have the same label (or, if multiple labels are present, -they should at least represent closely related cell states). +they should at least represent closely related cell states). We only +print out the top-left corner of the table here, but you should try +looking at the whole thing:
R
@@ -3024,231 +3026,23 @@R tab <- table(new.labels, sce$celltype.mapped) -tab +tab[1:4,1:4]
OUTPUT
+new.labels Cardiomyocytes + Allantois 0 + Cardiomyocytes 27 + Endothelium 0 + Erythroid2 0-new.labels Allantois Blood progenitors 1 - Allantois 28 0 - Cardiomyocytes 0 0 - Endothelium 0 0 - Erythroid2 0 0 - Erythroid3 0 0 - ExE mesoderm 0 0 - Forebrain/Midbrain/Hindbrain 1 0 - Gut 0 0 - Haematoendothelial progenitors 1 0 - Intermediate mesoderm 0 0 - Mesenchyme 0 0 - Neural crest 11 4 - NMP 0 0 - Paraxial mesoderm 5 0 - Pharyngeal mesoderm 0 0 - Somitic mesoderm 0 0 - Spinal cord 0 0 - Surface ectoderm 0 0 - -new.labels Blood progenitors 2 Cardiomyocytes - Allantois 0 0 - Cardiomyocytes 0 27 - Endothelium 0 0 - Erythroid2 1 0 - Erythroid3 0 0 - ExE mesoderm 0 0 - Forebrain/Midbrain/Hindbrain 0 0 - Gut 0 0 - Haematoendothelial progenitors 0 0 - Intermediate mesoderm 0 0 - Mesenchyme 0 0 - Neural crest 6 1 - NMP 0 0 - Paraxial mesoderm 0 4 - Pharyngeal mesoderm 0 6 - Somitic mesoderm 0 0 - Spinal cord 0 0 - Surface ectoderm 0 0 - -new.labels Caudal epiblast Caudal Mesoderm Def. endoderm - Allantois 0 0 0 - Cardiomyocytes 0 0 0 - Endothelium 0 0 0 - Erythroid2 0 0 0 - Erythroid3 0 0 0 - ExE mesoderm 0 0 0 - Forebrain/Midbrain/Hindbrain 1 0 2 - Gut 0 0 0 - Haematoendothelial progenitors 0 0 0 - Intermediate mesoderm 0 0 0 - Mesenchyme 0 0 0 - Neural crest 0 0 1 - NMP 0 0 0 - Paraxial mesoderm 0 0 1 - Pharyngeal mesoderm 0 0 0 - Somitic mesoderm 0 1 1 - Spinal cord 0 0 0 - Surface ectoderm 0 0 0 - -new.labels Doublet Endothelium Erythroid1 Erythroid2 - Allantois 0 0 0 0 - Cardiomyocytes 0 0 0 0 - Endothelium 0 13 0 0 - Erythroid2 0 0 14 21 - Erythroid3 0 0 6 5 - ExE mesoderm 0 0 0 0 - Forebrain/Midbrain/Hindbrain 6 0 0 0 - Gut 0 0 0 0 - Haematoendothelial progenitors 0 19 0 0 - Intermediate mesoderm 13 0 0 0 - Mesenchyme 0 0 0 0 - Neural crest 20 1 3 0 - NMP 0 0 0 0 - Paraxial mesoderm 5 0 0 0 - Pharyngeal mesoderm 1 0 0 0 - Somitic mesoderm 1 0 0 0 - Spinal cord 1 0 0 0 - Surface ectoderm 0 0 0 0 - -new.labels Erythroid3 ExE mesoderm - Allantois 0 0 - Cardiomyocytes 0 0 - Endothelium 0 0 - Erythroid2 51 0 - Erythroid3 49 0 - ExE mesoderm 0 23 - Forebrain/Midbrain/Hindbrain 0 1 - Gut 0 0 - Haematoendothelial progenitors 0 0 - Intermediate mesoderm 0 16 - Mesenchyme 0 0 - Neural crest 0 17 - NMP 0 0 - Paraxial mesoderm 0 0 - Pharyngeal mesoderm 0 0 - Somitic mesoderm 0 0 - Spinal cord 0 0 - Surface ectoderm 0 0 +new.labels Allantois Blood progenitors 1 Blood progenitors 2 + Allantois 28 0 0 + Cardiomyocytes 0 0 0 + Endothelium 0 0 0 + Erythroid2 0 0 1 -new.labels Forebrain/Midbrain/Hindbrain Gut - Allantois 0 0 - Cardiomyocytes 0 0 - Endothelium 0 0 - Erythroid2 0 0 - Erythroid3 0 0 - ExE mesoderm 0 0 - Forebrain/Midbrain/Hindbrain 83 4 - Gut 0 5 - Haematoendothelial progenitors 0 0 - Intermediate mesoderm 0 0 - Mesenchyme 0 0 - Neural crest 20 0 - NMP 0 0 - Paraxial mesoderm 0 0 - Pharyngeal mesoderm 0 0 - Somitic mesoderm 0 0 - Spinal cord 0 0 - Surface ectoderm 0 12 - -new.labels Haematoendothelial progenitors - Allantois 0 - Cardiomyocytes 0 - Endothelium 0 - Erythroid2 0 - Erythroid3 0 - ExE mesoderm 0 - Forebrain/Midbrain/Hindbrain 0 - Gut 0 - Haematoendothelial progenitors 16 - Intermediate mesoderm 0 - Mesenchyme 0 - Neural crest 8 - NMP 0 - Paraxial mesoderm 2 - Pharyngeal mesoderm 0 - Somitic mesoderm 0 - Spinal cord 0 - Surface ectoderm 0 - -new.labels Intermediate mesoderm Mesenchyme Neural crest - Allantois 0 0 0 - Cardiomyocytes 0 0 0 - Endothelium 0 0 0 - Erythroid2 0 0 0 - Erythroid3 0 0 0 - ExE mesoderm 1 7 0 - Forebrain/Midbrain/Hindbrain 0 3 0 - Gut 0 0 0 - Haematoendothelial progenitors 0 0 0 - Intermediate mesoderm 9 15 0 - Mesenchyme 0 69 0 - Neural crest 4 4 10 - NMP 0 0 0 - Paraxial mesoderm 0 7 0 - Pharyngeal mesoderm 0 13 0 - Somitic mesoderm 2 0 0 - Spinal cord 0 0 0 - Surface ectoderm 0 0 0 - -new.labels NMP Paraxial mesoderm Pharyngeal mesoderm - Allantois 0 0 0 - Cardiomyocytes 0 0 0 - Endothelium 0 0 0 - Erythroid2 0 0 0 - Erythroid3 0 0 0 - ExE mesoderm 0 0 0 - Forebrain/Midbrain/Hindbrain 41 2 4 - Gut 0 0 0 - Haematoendothelial progenitors 0 0 0 - Intermediate mesoderm 0 1 5 - Mesenchyme 0 0 0 - Neural crest 5 20 10 - NMP 16 0 0 - Paraxial mesoderm 0 43 3 - Pharyngeal mesoderm 0 0 18 - Somitic mesoderm 0 0 0 - Spinal cord 0 0 0 - Surface ectoderm 0 0 0 - -new.labels Rostral neurectoderm Somitic mesoderm - Allantois 0 0 - Cardiomyocytes 0 0 - Endothelium 0 0 - Erythroid2 0 0 - Erythroid3 0 0 - ExE mesoderm 0 0 - Forebrain/Midbrain/Hindbrain 11 1 - Gut 0 0 - Haematoendothelial progenitors 0 0 - Intermediate mesoderm 0 2 - Mesenchyme 0 0 - Neural crest 0 7 - NMP 0 0 - Paraxial mesoderm 0 0 - Pharyngeal mesoderm 0 0 - Somitic mesoderm 0 25 - Spinal cord 0 0 - Surface ectoderm 0 0 - -new.labels Spinal cord Stripped Surface ectoderm - Allantois 0 0 0 - Cardiomyocytes 0 0 0 - Endothelium 0 1 0 - Erythroid2 0 1 0 - Erythroid3 0 1 0 - ExE mesoderm 0 0 0 - Forebrain/Midbrain/Hindbrain 37 0 17 - Gut 0 0 2 - Haematoendothelial progenitors 0 0 0 - Intermediate mesoderm 0 0 0 - Mesenchyme 0 0 2 - Neural crest 3 5 6 - NMP 0 0 0 - Paraxial mesoderm 0 0 0 - Pharyngeal mesoderm 0 0 0 - Somitic mesoderm 0 0 0 - Spinal cord 9 0 0 - Surface ectoderm 0 0 20As a diagnostic measure, we examine the distribution of AUCs across cells for each label. In heterogeneous populations, the distribution for @@ -3292,7 +3086,7 @@
Challenge
-+The example that jumps out most strongly to the eye is ExE endoderm, which doesn’t show clear separate modes. Simultaneously, Endothelium @@ -3441,7 +3235,7 @@
Exercise 1: Clustering
-+The
NNGraphParamconstructor has an argumentcluster.args. This allows to specify arguments passed on to @@ -3458,7 +3252,7 @@Give me a hint
-+@@ -3487,7 +3281,7 @@TODO
Exercise 2: Cluster annotation
-+Use the
goana()function from the limma package to identify GO BP terms that are overrepresented in the list of @@ -3501,7 +3295,7 @@Give me a hint
-+@@ -3531,7 +3325,7 @@TODO
Exercise 3: Workflow
-+@@ -3603,7 +3397,7 @@TODO
Key Points
-->Content from Multi-sample analyses
-Last updated on 2024-09-06 | +
Last updated on 2024-09-08 | Edit this page
Estimated time: 45 minutes
@@ -3661,13 +3455,23 @@Objectives
replicate (pool 3, 4, and 5), we have both host and injected cells.We start by loading the data and doing a quick exploratory analysis, essentially applying the normalization and visualization techniques that -we have seen in the previous lectures to all samples.
+we have seen in the previous lectures to all samples. Note that this +time we’re selecting samples 5 to 10, not just 5 by itself.R
library(MouseGastrulationData) -sce <- WTChimeraData(samples=5:10, type = "processed") +library(batchelor) +library(edgeR) +library(scater) +library(ggplot2) +library(scran) +library(pheatmap) +library(scuttle) + +sce <- WTChimeraData(samples = 5:10, type = "processed") + sce@@ -3742,15 +3546,12 @@-OUTPUT<
R
-
library(scater) -library(ggplot2) -library(scran) +drop <- sce$celltype.mapped %in% c("stripped", "Doublet") -# remove doublets -drop <- sce$celltype.mapped %in% c("stripped", "Doublet") sce <- sce[,!drop] set.seed(29482) + idx <- unlist(tapply(colnames(sce), sce$sample, function(x) { perc <- round(0.50 * length(x)) sample(x, perc) @@ -3758,23 +3559,24 @@R sce <- sce[,idx]
We now normalize the data and visualize them in a tSNE plot.
+We now normalize the data, run some dimensionality reduction steps, +and visualize them in a tSNE plot.
R
-
# normalization -sce <- logNormCounts(sce) +sce <- logNormCounts(sce) + +dec <- modelGeneVar(sce, block = sce$sample) -# identify highly variable genes -dec <- modelGeneVar(sce, block=sce$sample) chosen.hvgs <- dec$bio > 0 -# dimensionality reduction sce <- runPCA(sce, subset_row = chosen.hvgs, ntop = 1000) + sce <- runTSNE(sce, dimred = "PCA") sce$sample <- as.factor(sce$sample) + plotTSNE(sce, colour_by = "sample")@@ -3810,7 +3612,7 @@Challenge
-+@@ -3831,29 +3633,30 @@Samples 5 and 6 were from the same “pool” of cells. Looking at the documentation for the dataset under
?WTChimeraDatawe see @@ -3822,7 +3624,7 @@Show me the solution
first TSNE plot. If you were developing single-cell library preparation protocols you might want to preserve this effect to understand how variation in pools leads to variation in expression, but for now, given -that we’re investigate other effects, we’ll want to remove this as +that we’re investigating other effects, we’ll want to remove this as undesired technical variation.Show me the solution
Correcting batch effects
-We correct the effect of samples by aid of the +
We “correct” the effect of samples with the
+package and using thecorrectExperimentfunction in thebatchelor-package and using thesamplecolDatacolumn as -batch.samplecolumn as batch.R
-
+merged <- runTSNE(merged, dimred = "corrected") + +plotTSNE(merged, colour_by = "batch")library(batchelor) -set.seed(10102) -merged <- correctExperiments(sce, - batch=sce$sample, - subset.row=chosen.hvgs, - PARAM=FastMnnParam( - merge.order=list( +set.seed(10102) + +merged <- correctExperiments( + sce, + batch = sce$sample, + subset.row = chosen.hvgs, + PARAM = FastMnnParam( + merge.order = list( list(1,3,5), # WT (3 replicates) list(2,4,6) # td-Tomato (3 replicates) ) ) ) -merged <- runTSNE(merged, dimred="corrected") -plotTSNE(merged, colour_by="batch")Once we removed the sample batch effect, we can proceed with the Differential Expression Analysis.
@@ -3875,7 +3678,7 @@Challenge
-+It’s important to have multiple samples within each experimental group because it helps the batch effect correction algorithm distinguish @@ -3898,7 +3701,7 @@
Show me the solution
Differential Expression
-In order to perform a Differential Expression Analysis, we need to +
In order to perform a differential expression analysis, we need to identify groups of cells across samples/conditions (depending on the experimental design and the final aim of the experiment).
As previously seen, we have two ways of grouping cells, cell @@ -3926,9 +3729,12 @@
R
# Using 'label' and 'sample' as our two factors; each column of the output # corresponds to one unique combination of these two factors. -library(scuttle) -summed <- aggregateAcrossCells(merged, - id=colData(merged)[,c("celltype.mapped", "sample")]) + +summed <- aggregateAcrossCells( + merged, + id = colData(merged)[,c("celltype.mapped", "sample")] +) + summed@@ -3965,11 +3771,11 @@R
label <- "Mesenchyme" -current <- summed[,label==summed$celltype.mapped] -# Creating up a DGEList object for use in edgeR: -library(edgeR) -y <- DGEList(counts(current), samples=colData(current)) +current <- summed[,label == summed$celltype.mapped] + +y <- DGEList(counts(current), samples = colData(current)) + y@@ -4018,7 +3824,9 @@R
discarded <- current$ncells < 10 + y <- y[,!discarded] + summary(discarded)@@ -4034,8 +3842,10 @@OUTPUT<
R
-
keep <- filterByExpr(y, group=current$tomato) +keep <- filterByExpr(y, group = current$tomato) + y <- y[keep,] + summary(keep)@@ -4055,6 +3865,7 @@R
y <- calcNormFactors(y) + y$samples@@ -4095,7 +3906,7 @@Rpar(mfrow = c(2,3)) for (i in seq_len(ncol(y))) { - plotMD(y, column=i) + plotMD(y, column = i) }
@@ -4112,8 +3923,8 @@RR
-
+limma::plotMDS(cpm(y, log=TRUE), - col=ifelse(y$samples$tomato, "red", "blue"))limma::plotMDS(cpm(y, log = TRUE), + col = ifelse(y$samples$tomato, "red", "blue"))We then construct a design matrix by including both the pool and the tomato as factors. This design indicates which samples belong to which @@ -4153,6 +3964,7 @@
R
y <- estimateDisp(y, design) + summary(y$trended.dispersion)@@ -4172,7 +3984,7 @@RplotBCV(y)
We then fit a Quasi-Likelihood (QL) negative binomial generalized -linear model for each gene. The
robust=TRUEparameter +linear model for each gene. Therobust = TRUEparameter avoids distortions from highly variable clusters. The QL method includes an additional dispersion parameter, useful to handle the uncertainty and variability of the per-gene variance, which is not well estimated by the @@ -4182,7 +3994,8 @@RR
-
fit <- glmQLFit(y, design, robust=TRUE) +fit <- glmQLFit(y, design, robust = TRUE) + summary(fit$var.prior)@@ -4221,7 +4034,8 @@RR
-
res <- glmQLFTest(fit, coef=ncol(design)) +res <- glmQLFTest(fit, coef = ncol(design)) + summary(decideTests(res))@@ -4261,14 +4075,14 @@OUTPUT<
R
-
library(scran) -summed.filt <- summed[,summed$ncells >= 10] +summed.filt <- summed[,summed$ncells >= 10] -de.results <- pseudoBulkDGE(summed.filt, - label=summed.filt$celltype.mapped, - design=~factor(pool) + tomato, - coef="tomatoTRUE", - condition=summed.filt$tomato +de.results <- pseudoBulkDGE( + summed.filt, + label = summed.filt$celltype.mapped, + design = ~factor(pool) + tomato, + coef = "tomatoTRUE", + condition = summed.filt$tomato )The returned object is a list of
DataFrames each with @@ -4280,6 +4094,7 @@R
cur.results <- de.results[["Allantois"]] + cur.results[order(cur.results$PValue),]@@ -4317,7 +4132,7 @@Challenge
-+“logFC” stands for log fold-change. Rather than reporting e.g. a 5-fold increase, it’s better to report a logFC of log(5) = 1.61. @@ -4336,26 +4151,29 @@
Show me the solution
With DA we look for differences in cluster abundance across conditions (the tomato injection in our case), rather than differences in gene expression.
-We first setup some code and variables for further analysis, like -quantifying the number of cells per each cell type and fit a model to -catch differences between the injected cells and the background.
-The steps are very similar to the ones for DEGs analysis, but this -time we start our analysis on the computed abundances and without +
Our first steps are quantifying the number of cells per each cell +type and fitting a model to catch differences between the injected cells +and the background.
+The process is very similar differential expression modeling, but +this time we start our analysis on the computed abundances and without normalizing the data with TMM.
R
-
+ +y.ab <- estimateDisp(y.ab, design, trend = "none") + +fit.ab <- glmQLFit(y.ab, design, robust = TRUE, abundance.trend = FALSE)library(edgeR) -abundances <- table(merged$celltype.mapped, merged$sample) +abundances <- table(merged$celltype.mapped, merged$sample) + abundances <- unclass(abundances) -# Attaching some column metadata. + extra.info <- colData(merged)[match(colnames(abundances), merged$sample),] -y.ab <- DGEList(abundances, samples=extra.info) + +y.ab <- DGEList(abundances, samples = extra.info) design <- model.matrix(~factor(pool) + factor(tomato), y.ab$samples) -y.ab <- estimateDisp(y.ab, design, trend="none") -fit.ab <- glmQLFit(y.ab, design, robust=TRUE, abundance.trend=FALSE)Background on compositional effect @@ -4401,6 +4219,7 @@
R
y.ab2 <- calcNormFactors(y.ab) + y.ab2$samples$norm.factors@@ -4413,11 +4232,11 @@@@ -4433,7 +4252,7 @@OUTPUT<
R
-
y.ab2 <- estimateDisp(y.ab2, design, trend="none") +y.ab2 <- estimateDisp(y.ab2, design, trend = "none") -fit.ab2 <- glmQLFit(y.ab2, design, robust=TRUE, abundance.trend=FALSE) +fit.ab2 <- glmQLFit(y.ab2, design, robust = TRUE, abundance.trend = FALSE) -res2 <- glmQLFTest(fit.ab2, coef=ncol(design)) +res2 <- glmQLFTest(fit.ab2, coef = ncol(design)) summary(decideTests(res2))OUTPUT<
R
-
+topTags(res2, n=10)topTags(res2, n = 10)OUTPUT @@ -4469,7 +4288,8 @@
Testing against a log-fold
R
-
res.lfc <- glmTreat(fit.ab, coef=ncol(design), lfc=1) +res.lfc <- glmTreat(fit.ab, coef = ncol(design), lfc = 1) + summary(decideTests(res.lfc))@@ -4550,70 +4370,72 @@OUTPUT< [8] base other attached packages: - [1] edgeR_4.2.0 limma_3.60.2 - [3] batchelor_1.20.0 scran_1.32.0 - [5] scater_1.32.0 ggplot2_3.5.1 - [7] scuttle_1.14.0 MouseGastrulationData_1.18.0 - [9] SpatialExperiment_1.14.0 SingleCellExperiment_1.26.0 -[11] SummarizedExperiment_1.34.0 Biobase_2.64.0 -[13] GenomicRanges_1.56.0 GenomeInfoDb_1.40.1 -[15] IRanges_2.38.0 S4Vectors_0.42.0 -[17] BiocGenerics_0.50.0 MatrixGenerics_1.16.0 -[19] matrixStats_1.3.0 BiocStyle_2.32.0 + [1] pheatmap_1.0.12 scran_1.32.0 + [3] scater_1.32.0 ggplot2_3.5.1 + [5] scuttle_1.14.0 edgeR_4.2.0 + [7] limma_3.60.2 batchelor_1.20.0 + [9] MouseGastrulationData_1.18.0 SpatialExperiment_1.14.0 +[11] SingleCellExperiment_1.26.0 SummarizedExperiment_1.34.0 +[13] Biobase_2.64.0 GenomicRanges_1.56.0 +[15] GenomeInfoDb_1.40.1 IRanges_2.38.0 +[17] S4Vectors_0.42.0 BiocGenerics_0.50.0 +[19] MatrixGenerics_1.16.0 matrixStats_1.3.0 +[21] BiocStyle_2.32.0 loaded via a namespace (and not attached): - [1] DBI_1.2.3 formatR_1.14 - [3] gridExtra_2.3 rlang_1.1.3 - [5] magrittr_2.0.3 compiler_4.4.1 - [7] RSQLite_2.3.7 DelayedMatrixStats_1.26.0 - [9] png_0.1-8 vctrs_0.6.5 - [11] pkgconfig_2.0.3 crayon_1.5.2 - [13] fastmap_1.2.0 dbplyr_2.5.0 - [15] magick_2.8.3 XVector_0.44.0 - [17] labeling_0.4.3 utf8_1.2.4 - [19] rmarkdown_2.27 UCSC.utils_1.0.0 - [21] ggbeeswarm_0.7.2 purrr_1.0.2 - [23] bit_4.0.5 bluster_1.14.0 - [25] xfun_0.44 zlibbioc_1.50.0 - [27] cachem_1.1.0 beachmat_2.20.0 - [29] jsonlite_1.8.8 blob_1.2.4 - [31] highr_0.11 DelayedArray_0.30.1 - [33] BiocParallel_1.38.0 cluster_2.1.6 - [35] irlba_2.3.5.1 parallel_4.4.1 - [37] R6_2.5.1 Rcpp_1.0.12 - [39] knitr_1.47 splines_4.4.1 - [41] igraph_2.0.3 Matrix_1.7-0 - [43] tidyselect_1.2.1 viridis_0.6.5 - [45] abind_1.4-5 yaml_2.3.8 - [47] codetools_0.2-20 curl_5.2.1 - [49] lattice_0.22-6 tibble_3.2.1 - [51] withr_3.0.0 KEGGREST_1.44.0 - [53] BumpyMatrix_1.12.0 Rtsne_0.17 - [55] evaluate_0.23 BiocFileCache_2.12.0 - [57] ExperimentHub_2.12.0 Biostrings_2.72.1 - [59] pillar_1.9.0 BiocManager_1.30.23 - [61] filelock_1.0.3 renv_1.0.7 - [63] generics_0.1.3 BiocVersion_3.19.1 - [65] sparseMatrixStats_1.16.0 munsell_0.5.1 - [67] scales_1.3.0 glue_1.7.0 - [69] metapod_1.12.0 tools_4.4.1 - [71] AnnotationHub_3.12.0 BiocNeighbors_1.22.0 - [73] ScaledMatrix_1.12.0 locfit_1.5-9.9 - [75] cowplot_1.1.3 grid_4.4.1 - [77] AnnotationDbi_1.66.0 colorspace_2.1-0 - [79] GenomeInfoDbData_1.2.12 beeswarm_0.4.0 - [81] BiocSingular_1.20.0 vipor_0.4.7 - [83] cli_3.6.2 rsvd_1.0.5 - [85] rappdirs_0.3.3 fansi_1.0.6 - [87] viridisLite_0.4.2 S4Arrays_1.4.1 - [89] dplyr_1.1.4 ResidualMatrix_1.14.0 - [91] gtable_0.3.5 digest_0.6.35 - [93] dqrng_0.4.1 SparseArray_1.4.8 - [95] ggrepel_0.9.5 farver_2.1.2 - [97] rjson_0.2.21 memoise_2.0.1 - [99] htmltools_0.5.8.1 lifecycle_1.0.4 -[101] httr_1.4.7 statmod_1.5.0 -[103] mime_0.12 bit64_4.0.5 + [1] RColorBrewer_1.1-3 jsonlite_1.8.8 + [3] magrittr_2.0.3 ggbeeswarm_0.7.2 + [5] magick_2.8.3 farver_2.1.2 + [7] rmarkdown_2.27 zlibbioc_1.50.0 + [9] vctrs_0.6.5 memoise_2.0.1 + [11] DelayedMatrixStats_1.26.0 htmltools_0.5.8.1 + [13] S4Arrays_1.4.1 AnnotationHub_3.12.0 + [15] curl_5.2.1 BiocNeighbors_1.22.0 + [17] SparseArray_1.4.8 cachem_1.1.0 + [19] ResidualMatrix_1.14.0 igraph_2.0.3 + [21] mime_0.12 lifecycle_1.0.4 + [23] pkgconfig_2.0.3 rsvd_1.0.5 + [25] Matrix_1.7-0 R6_2.5.1 + [27] fastmap_1.2.0 GenomeInfoDbData_1.2.12 + [29] digest_0.6.35 colorspace_2.1-0 + [31] AnnotationDbi_1.66.0 dqrng_0.4.1 + [33] irlba_2.3.5.1 ExperimentHub_2.12.0 + [35] RSQLite_2.3.7 beachmat_2.20.0 + [37] filelock_1.0.3 labeling_0.4.3 + [39] fansi_1.0.6 httr_1.4.7 + [41] abind_1.4-5 compiler_4.4.1 + [43] bit64_4.0.5 withr_3.0.0 + [45] BiocParallel_1.38.0 viridis_0.6.5 + [47] DBI_1.2.3 highr_0.11 + [49] rappdirs_0.3.3 DelayedArray_0.30.1 + [51] rjson_0.2.21 bluster_1.14.0 + [53] tools_4.4.1 vipor_0.4.7 + [55] beeswarm_0.4.0 glue_1.7.0 + [57] grid_4.4.1 Rtsne_0.17 + [59] cluster_2.1.6 generics_0.1.3 + [61] gtable_0.3.5 BiocSingular_1.20.0 + [63] ScaledMatrix_1.12.0 metapod_1.12.0 + [65] utf8_1.2.4 XVector_0.44.0 + [67] ggrepel_0.9.5 BiocVersion_3.19.1 + [69] pillar_1.9.0 BumpyMatrix_1.12.0 + [71] splines_4.4.1 dplyr_1.1.4 + [73] BiocFileCache_2.12.0 lattice_0.22-6 + [75] renv_1.0.7 bit_4.0.5 + [77] tidyselect_1.2.1 locfit_1.5-9.9 + [79] Biostrings_2.72.1 knitr_1.47 + [81] gridExtra_2.3 xfun_0.44 + [83] statmod_1.5.0 UCSC.utils_1.0.0 + [85] yaml_2.3.8 evaluate_0.23 + [87] codetools_0.2-20 tibble_3.2.1 + [89] BiocManager_1.30.23 cli_3.6.2 + [91] munsell_0.5.1 Rcpp_1.0.12 + [93] dbplyr_2.5.0 png_0.1-8 + [95] parallel_4.4.1 blob_1.2.4 + [97] sparseMatrixStats_1.16.0 viridisLite_0.4.2 + [99] scales_1.3.0 purrr_1.0.2 +[101] crayon_1.5.2 rlang_1.1.3 +[103] cowplot_1.1.3 KEGGREST_1.44.0 +[105] formatR_1.14
Exercises
@@ -4636,7 +4458,7 @@Exercise 1: Replicates
-+Remember, you can subset SingleCellExperiments with logical indices, just like a matrix. You can also access their column data with the @@ -4650,7 +4472,7 @@
Give me a hint
-+@@ -4686,7 +4509,7 @@R @@ -4658,11 +4480,12 @@
R
summed.filt.subset = summed.filt[,summed.filt$pool != 3] -de.results <- pseudoBulkDGE(summed.filt.subset, - label=summed.filt.subset$celltype.mapped, - design=~factor(pool) + tomato, - coef="tomatoTRUE", - condition=summed.filt.subset$tomato +de.results <- pseudoBulkDGE( + summed.filt.subset, + label = summed.filt.subset$celltype.mapped, + design = ~factor(pool) + tomato, + coef = "tomatoTRUE", + condition = summed.filt.subset$tomato )Exercise 2:
-+You can just hand
pheatmap()a matrix as its only argument. It has a million options, but the defaults are usually pretty @@ -4700,15 +4523,13 @@Give me a hint
-+R
-
+library(pheatmap) - -pheatmap(y.ab$counts)pheatmap(y.ab$counts)The top DA result was a decrease in ExE ectoderm in the tomato condition, which you can sort of see, especially if you @@ -6470,7 +6291,7 @@
Exercise 1
-+R @@ -6503,7 +6324,7 @@
Exercise 2
-+R @@ -6535,7 +6356,7 @@
Exercise 3
-+
Install R using your package manager and RStudio
-+
If you already have R and RStudio installed
-+