scmap-scripts

This is a collection of R scripts to allow workflow-driven execution of differnt steps of the scmap workflow.

Commands

Currently wrapped scmap functions are described below. Each script has usage insructions available via --help, consult function documentation in scmap for further details.

Preprocess SCE object for scmap pipeline

This script makes the necessary changes to the SCE object required by the scmap workflow, including 'un-sparsing' and log-normalising the expression matrix.

scmap-preprocess-sce.R --input-object <path to the SCE object>\
                       --output-sce-object <path to the updated SCE object in .rds format>

Extract test data

Input to the workflow will be a serialised single-cell experiment object. You can generate one for testing (derived from the package-provided test data) like:

scmap-make-test-data.R --output-object-file <output SingleCellExperiment in .rds format>

Find the most informative features (genes/transcripts) for projection

scmap-select-features.R --input-object-file <input SingleCellExperiment in .rds format>  \
    --n-features <number features to use> --output-object-file <output SingleCellExperiment in .rds format> \
    --output-plot-file <optional file name in .png format, for feature selection plot>

Calculate centroids of each cell type and merge them into a single table.

Here we generate a summary representation of each cluster in the indexed dataset:

scmap-index-cluster.R --input-object-file <input SingleCellExperiment in .rds format> \
     --cluster-col <column name where cell types are stored> \
     --train-id <Training dataset ID (optional)> \
     --remove-mat <Should expression data be removed from the index? >\
     --output-object-file <output SingleCellExperiment in .rds format> \
     --output-plot-file <optional file name in .png format, for heatmap-style index visualisation>

Project one dataset to another

In this step we find the cluster medoid of the index dataset closest to the cells of a query:

scmap-scmap-cluster.R -i <cluster-indexed SingleCellExperiment in .rds format> \
    -p <query SingleCellExperiment in .rds format> --threshold <cluster similarity threshold> \
    --output-text-file <csv-format file to store results> \
    --output-object-file <output SingleCellExperiment in .rds format>

Create an index for a dataset to enable fast approximate nearest neighbour search

Here we generate a cell-wise index:

scmap-index-cell.R --input-object-file <input SingleCellExperiment in .rds format> \
    --train-id <Training dataset ID (optional)> \
    --number-chunks <number of chunks into which the expr matrix is split> \
    --remove-mat <Should expression data be removed from the index? >\
    --number-clusters <number of clusters per group for k-means clustering> \
    --output-object-file <output SingleCellExperiment in .rds format>

For each cell in a query dataset, search for the nearest neighbours by cosine distance within a collection of reference datasets

Here we find the nearest 'n' neighbours in an index dataset for the cells of a query dataset. Optionally (when --cluster-col is set and corresponds to a column in the index dataset's colData()), generate a cluster identity for query cells via the cluster inenties in the index:

scmap-scmap-cell.R -i $index_cell_sce -p <input SingleCellExperiment in .rds format> \
    --number-nearest-neighbours <number nearest neighbours> \
    --cluster-col <column name where cell types are stored> \
    --output-object-file <output SingleCellExperiment in .rds format> \
    --output-clusters-text-file <file to store optional cluster identities> \
    --closest-cells-text-file <csv file to store closest cells> \
    --closest-cells-similarities-text-file <csv file to store similarity values>

Get standard output for downstream processing and analysis as part of various workflows

scmap_get_std_output.R\
            --predictions-file <Path to the predictions file in text format>\
            --output-table <Path to the final output file in text format>\
            --include-scores <Boolean: Should prediction scores be included in output? Default: FALSE>\
            --tool <What tool produced output? (scmap-cell or scmap-cluster)>\
            --index <Path to the index object in .rds format (Optional; required to add dataset of origin to output table)>\
            --sim-col-name <Column name of similarity scores>