-
Notifications
You must be signed in to change notification settings - Fork 25
/
Copy pathbsdata_glm2.R
executable file
·165 lines (134 loc) · 6.8 KB
/
bsdata_glm2.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
#!/usr/bin/env Rscript
suppressPackageStartupMessages(library("argparse"))
description<- 'Read matrix files for chrom and produce GLM output. Example bsdata_glm2.R -d design.txt -c BS oxBS -i chr1 -m 5 -M 1000 -l 4'
# create parser object
parser <- ArgumentParser(description= description)
parser$add_argument("-i", "--indir", required= TRUE,
help="Directory with input files")
parser$add_argument("-b", "--basename", default= 'genome',
help="Basename to find input files inside --indir. Default 'genome'")
parser$add_argument("-d", "--design", required= TRUE,
help="File with the experimental design")
parser$add_argument("-c", "--contrast", required= TRUE,
nargs= 2,
help= "Contrasts to apply")
parser$add_argument("-m", "--minCov", default= 0,
type= "integer",
help= "Minimum coverage for a position to be retained")
parser$add_argument("-M", "--maxCov", default= 1000000,
type= "integer",
help= "Max coverage for a position to be retained")
parser$add_argument("-l", "--minLibs", default= 1,
type= "integer",
help= "Min number of libraries that must satisfy -m and -M. Default 1: It's enough that one lib has >= minCov and <= maxCov for the locus to be included.")
parser$add_argument("-n", "--nprocs", default= 1,
type= "integer",
help= "Number of processors to use")
parser$add_argument("--nrows", default= -1,
type= "integer",
help= "For debugging: Number of rows from input to process. Default -1 (all rows)")
args <- parser$parse_args()
# ------------------------- Setting up ----------------------------------------
source('/home/berald01/svn_checkout/bioinformatics-misc/glm.R')
source('/home/berald01/svn_checkout/bioinformatics-misc/BSdata.R')
source('/home/berald01/svn_checkout/bioinformatics-misc/makeTransparent.R')
# -----------------------------------------------------------------------------
## Read design
## -----------
design<- read.table(args$design, header= TRUE, sep= '\t')
if(('library_id' %in% names(design)) == FALSE){
stop('design file must contain a column "library_id"')
}
if(('bs' %in% names(design)) == FALSE){
stop('\n\ndesign file must contain a column "bs"')
}
if(!all(args$contrast %in% design$bs)){
stop('\nInvalid contrasts')
}
if(length(unique(args$contrast)) != 2){
stop('Invalid number of contrasts')
}
if(args$minLibs > nrow(design)){
stop('More libraries in minLibs than available.')
}
## Read data
## ---------
cpg<- read.bsdata(file.path(args$indir, args$basename), mat= c('loci', 'cnt_met', 'tot_reads'), gzip= FALSE, nrows= args$nrows)
if(all(as.character(design$library_id) == as.character(cpg@design$library_id)) == FALSE){
print(design)
stop('\n\nLibraries IDs in design file do not match those in input files')
}
cpg@design$bs<- design$bs
cpg@pct_met<- as.ffdf(100 * (cpg@cnt_met[,] / cpg@tot_reads[,]))
# Read count
# ----------
cat('Plot of read counts...\n')
pdf(file.path(args$indir, 'tot_reads.hist.pdf'), width= 15/2.54, height= 12/2.54, pointsize= 9)
par(mfrow= c(2, 3), mar= c(2.5, 3, 2, 0.2), oma= c(3,3,2,0))
for(x in colnames(cpg@tot_reads)){
readQ<- quantile(cpg@tot_reads[, x], 0.95)
y<- ifelse(cpg@tot_reads[, x] > readQ, readQ, cpg@tot_reads[, x] )
hist(y, main= '', col= 'grey80', xlab= '', ylab= '', breaks= 10)
mtext(side= 3, line= -1, text= x, cex= 0.8)
}
mtext(side= 1, text= 'Read depth', font= 2, outer= TRUE)
mtext(side= 2, text= 'Frequency', outer= TRUE, font= 2)
mtext(side= 3, text= sprintf('%s Depth of coverage at CpG sites', args$indir), outer= TRUE, font= 2, line= 0)
dev.off()
# Methylation
# -----------
cat('Plot of methylation...\n')
## Select CpG with coverage n<x<m
m<- 5
Q<- 0.95
pdf(file.path(args$indir, 'pct_met.hist.pdf'), width= 15/2.54, height= 12/2.54, pointsize= 9)
par(mfrow= c(2, 3), mar= c(2.5, 3, 2, 0.2), oma= c(3,3,2,0))
for(x in colnames(cpg@pct_met)){
y<- cpg@pct_met[, x]
tot<- cpg@tot_reads[, x]
readQ<- quantile(cpg@tot_reads[, x], Q)
y<- y[which(tot >= m & tot <= readQ)]
hist(y, main= '', col= 'grey80', xlab= '', ylab= '', breaks= 25)
mtext(side= 3, line= -1, text= paste(x, '\nn= ', length(y), sep= ''), cex= 0.8, col= 'darkblue')
}
mtext(side= 1, text= '% methylated', font= 2, outer= TRUE)
mtext(side= 2, text= 'Frequency', outer= TRUE, font= 2)
mtext(side= 3, text= sprintf('%s CpG sites with depth %sx to quantile(%s)', args$indir, m, Q), outer= TRUE, font= 2, line= 0)
dev.off()
# Filter positions with not-too-high coverage
# -------------------------------------------
# Select correct columns
cpg.2<- BSdataApply(cpg, FUN= function(x) x[, which(cpg@design$bs %in% args$contrast)])
cpg.2@design<- cpg@design[which(cpg@design$bs %in% args$contrast),]
# Filter for coverage
filter<- which(rowSums(cpg.2@tot_reads[,] >= args$minCov & cpg.2@tot_reads[,] <= args$maxCov) >= args$minLibs) ## Positions with coverage x in at least y libs
cat(sprintf('Applying GLM to %s loci\n', length(filter)))
cpg.2<- BSdataApply(cpg.2, FUN= function(x) x[filter,])
pdf(file.path(args$indir, 'hist_depth_comb_libs.pdf'), width= 18/2.54, height= 10/2.54, pointsize= 10)
par(mfrow= c(1,2))
hist(rowSums(cpg.2@tot_reads[,]),
main= sprintf('Depth of coverage summed across %s libraries\nn= %s', ncol(cpg.2@tot_reads), nrow(cpg.2@tot_reads)),
xlab= 'Cumulated depth of coverage', ylab= 'No. CpG', breaks= 20)
hist(rowMeans(cpg.2@pct_met[,]),
main= sprintf('%s Methylation averaged across %s libraries\nn= %s', args$indir, ncol(cpg.2@tot_reads), nrow(cpg.2@tot_reads)),
xlab= '% methylation', ylab= 'No. CpG', breaks= 20)
dev.off()
# GLM
# ---
glmbs<- glmBS.2(bsobj= cpg.2, bs= cpg.2@design$bs, contrast= args$contrast, family= 'binomial', nthreads= args$nprocs)
row.names(glmbs)<- row.names(cpg.2@cnt_met)
write.table(glmbs, file= file.path(args$indir, 'glmbs.txt'), row.names= FALSE, col.names= TRUE, sep= '\t', quote= FALSE)
## MEMO: Estimate is "oxBS - BS" (reverse alphanumeric order)
pneg<- glmbs$pvalue[which(glmbs$estimate > 0)] ## Pvalues of negative estimates.
ppos<- glmbs$pvalue[which(glmbs$estimate < 0)]
pdf(file.path(args$indir, 'hist_pvals_glm.pdf'), width= 18/2.54, height= 10/2.54, pointsize= 8)
par(las= 0, mfrow= c(1, 2))
h1<- hist(pneg, border= 'transparent', add= FALSE, col= 'grey70', freq= FALSE, breaks= 20, xlab= 'P-value', main= "P-values from GLM binom")
h2<- hist(ppos, border= 'red', freq= FALSE, add= TRUE, breaks= 20)
points(pch= 17, col= 'blue', x= 0.05, y= 0)
h1<- hist(pneg, border= 'transparent', add= FALSE, col= 'grey70', freq= TRUE, breaks= 20, xlab= 'P-value', main= sprintf("%s P-values from GLM binom\nn= %s", args$indir, nrow(glmbs)))
h2<- hist(ppos, border= 'red', freq= TRUE, add= TRUE, breaks= 20)
points(pch= 17, col= 'blue', x= 0.05, y= 0)
legend('topleft', legend=c('Pval +ve %', 'Pval -ve %'), col= c('red', 'grey70'), pch= 15, bg= 'white')
dev.off()
quit(save= 'no')