-
Notifications
You must be signed in to change notification settings - Fork 0
/
deseq2_run_LR.R
271 lines (227 loc) · 11.9 KB
/
deseq2_run_LR.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
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
#!/usr/bin/env Rscript
library(getopt)
spec = matrix(c(
'bmF', 'B', 0, "logical", "Disable baseMean filtering. (False)",
'baseDir', 'b', 1, "character", "Base directory for everything (.)",
'lfcthreshold', 'c', 1, "numeric", "Log2 fold-change threshold (1)",
'fc4p', 'C', 0, "logical", "Incorporate the LFC threshold in the Null Hypothesis.",
'createID', 'D', 0, "logical", "Non-summarized intron or exon counts by featureCounts. The first column is not unique IDs because genes are repeated.",
'countsFile', 'f', 1, "character", "Tab-separated table of raw counts with `row_ID` and all the samples.",
'fullFormula', 'F', 1, "character", "Full model formula for LR or Wald test",
'specnorm', 'G', 1, "character", "Special normalisation. Gene ids that match this pattern will not be included for calculation of library sizes for TPM/RPM normalisation.",
'help', 'h', 0, "logical", "This help.",
'altHypo', 'H', 1, "character", "Alternative hypothesis type (greaterAbs).",
'idcol', 'i', 1, "numeric", "ID column to use (1). The others will be removed. See also -I.",
'nidcols', 'I', 1, "numeric", "Number of ID columns at the start of the table (1). See also -i.",
'prescaled', 'k', 0, "logical", "Don't let Deseq2 rescale the libraries. (False)",
'label', 'l', 0, "logical", "Add comparison details to the standard column names. Useful for merging multiple outputs.",
'minCount', 'm', 1, "numeric", "Minimum number of reads in either condition in a contrast. (100)",
'minTPM', 'M', 1, "numeric", "Minimum TPM in either condition in a contrast. (5)",
'ntop', 'n', 1, "numeric", "Number of hits to highlight (50)",
'resultsDir', 'o', 1, "character", "Directory in which to save the contrast results. If omitted, only the RDS will be output.",
'pcutoff', 'p', 1, "numeric", "P-value cutoff (0.05)",
'prefix', 'P', 1, "character", "Prefix to add to output file name and column names (''). ",
'reducedFormula','R', 1, "character", "Reduced formula for LR test.",
'samplesFile', 's', 1, "character", "Tab-separated table with `sample` column followed by the variable columns. Values must NOT contain punctuation other than '_' !!!",
'reportTemplate','T', 1, "character", "Template Rmd file for DE report.",
'vst', 'v', 0, "logical", "Plot VST counts instead of rlog2",
'widthsFile', 'w', 1, "character", "Full path to table of feature lengths",
'widthsCol', 'W', 1, "integer", "Column in the counts file that contains feature lengths. This should be among the non-count columns, see also -I. It is an alternative to providing a separate lengths file.",
'comparisons', 'x', 1, "character", "Comma separated list indicating which levels of the formula factor to compare, in order of appearance. ie 'Treatment-2v1,Treatment-3v1,Treatment-3v2,Batch-4v5'.",
'contexts', 'X', 1, "character", "Comma separated list indicating which levels of which covariate to use as context/subset for each comparison, ie 'celltype-1,celltype-1,celltype-2,culture-7"
), byrow=TRUE, ncol=5)
# opt <- list(createID=FALSE,
# baseDir='/SCRATCH/PP2023011_SLC13A5',
# countsFile='outputs/star_salmon/salmon.merged.gene_counts_length_scaled.tsv',
# resultsDir='outputs/diffex2.1',
# samplesFile='samplesheet2.differentialabundance.csv',
# minCount=10,
# minTPM=0,
# lfcthreshold=0.5,
# nidcols=2,
# idcol=2,
# ntop=50,
# bmF=FALSE,
# pcutoff=0.05,
# prescaled=FALSE,
# prefix='TESS',
# label=TRUE,
# widthsCol=NULL,
# reportTemplate="~/utility_scripts/deseq2_report_template_LR.Rmd",
# comparisons="condition-2v1,time2-2v1",
# contexts="binary-1,binary-1",
# fullFormula="~ time2 + condition",
# reducedFormula="NULL",
# specnorm="NULL")
opt = getopt(spec)
if ( !is.null(opt$help) ) {
cat(getopt(spec, usage=TRUE))
q(status=1)
}
library(DESeq2)
library(data.table)
stopifnot(!is.null(opt$countsFile))
stopifnot(!is.null(opt$samplesFile))
if (is.null(opt$reportTemplate)) opt$reportTemplate <- '~/utility_scripts/deseq2_report_template_LR.Rmd'
if (is.null(opt$createID)) opt$createID <- FALSE
if (!is.null(opt$specnorm) && opt$specnorm == "NULL") opt$specnorm <- NULL
if (is.null(opt$prescaled)) opt$prescaled <- FALSE
if (is.null(opt$label)) opt$label <- FALSE
if (is.null(opt$baseDir)) opt$baseDir <- '.'
if (is.null(opt$resultsDir)) opt$resultsDir <- './results'
if (is.null(opt$minCount)) opt$minCount <- 100
if (is.null(opt$minTPM)) opt$minTPM <- 5
if (is.null(opt$ntop)) opt$ntop <- 50
if (is.null(opt$pcutoff)) opt$pcutoff <- 0.05
if (is.null(opt$lfcthreshold)) opt$lfcthreshold <- 1
if (is.null(opt$altHypo)) opt$altHypo <- 'greaterAbs'
if (is.null(opt$nidcols)) opt$nidcols <- 1
if (is.null(opt$idcol)) opt$idcol <- 1
if (is.null(opt$bmF)) opt$bmF <- FALSE
opt$bmF <- !opt$bmF # Convert -B flag functionality from switch-on to switch-off.
if (is.null(opt$vst)) opt$vst <- FALSE
if (!is.null(opt$fullFormula) && opt$fullFormula == "NULL") opt$fullFormula <- NULL
if (!is.null(opt$reducedFormula) && opt$reducedFormula == "NULL") opt$reducedFormula <- NULL
if (!is.null(opt$comparisons) && opt$comparisons == "NULL") opt$comparisons <- NULL
if (is.null(opt$contexts) | opt$contexts == "NULL") stop("Need at least one context!")
# Output destinations
dir.create(file.path(opt$baseDir, opt$resultsDir, opt$prefix), recursive=TRUE)
# Covariates
covars <- read.csv(file.path(opt$baseDir, opt$samplesFile), sep=",", row.names="sample", header=TRUE, check.names = FALSE, colClasses="character")
# Contrasts
contexts <- strsplit(gsub(' ', '', opt$contexts), ',')[[1]]
contexts <- strsplit(contexts , '-')
if (!is.null(opt$comparisons)) {
comparisons <- strsplit(gsub(' ', '', opt$comparisons), ',')[[1]]
stopifnot(length(contexts) == length(comparisons))
comparisons <- strsplit(strsplit(opt$comparisons, ',')[[1]], '-')
comparisons <- lapply(comparisons, function(x){
# x <- comparisons[[1]]
list( x[1], as.integer(strsplit(x[2], 'v')[[1]]) )
})
}
# Counts
cts <- fread(file.path(opt$baseDir, opt$countsFile), sep="\t", header=TRUE, check.names = FALSE, colClasses="character")
# Setting all columns to character prevents numeric IDs at the top of the ID columns from auto-defining these columns as numeric when there could be character IDs further down that would otherwise become NaN.
cts <- unique(cts) # non-summarized featureCounts have repeated exons from sharing among transcripts.
# Remove disruptive features (such as immunoglobins in plasmablasts).
ogn <- nrow(cts)
if (!is.null(opt$specnorm))
cts <- cts[!grepl(opt$specnorm, Geneid), ]
deltan <- ogn - nrow(cts)
xref <- cts[, 1:opt$nidcols]
cts <- as.matrix(cts[, rownames(covars), with=FALSE]) # Order and subset columns
cts <- matrix(as.numeric(cts), ncol=ncol(cts))
colnames(cts) <- row.names(covars) # because the type conversion lost them
cts[is.na(cts)] <- 0
if (opt$createID) {
xref[, id := paste0(Geneid, '.', Start, '_', End)]
if (any(duplicated(xref$id)))
stop()
rownames(cts) <- xref$id
} else {
rownames(cts) <- xref[[opt$idcol]]
}
# Feature lengths
if (!is.null(opt$widthsCol)) {
featLens <- data.table(ID=rownames(cts),
LEN=as.integer(xref[[opt$widthsCol]]))
} else if (!is.null(opt$widthsFile)) {
featLens <- fread(opt$widthsFile, sep="\t", header=TRUE, check.names = FALSE, colClasses=c("character", "numeric"))
# Match order of feature lengths to order of features
setkeyv(featLens, names(featLens)[1])
featLens <- featLens[rownames(cts), ]
} else {
featLens <- NULL
}
stopifnot(all(rownames(cts) == featLens[[1]]))
#######################
#######################
### Split up data as instructed
for(y in unique(contexts)) {
# y <- unique(contexts)[[1]]
contvar <- y[1]
contlev <- unique(covars[[contvar]])[as.integer(y[2])]
sel <- which(vapply(contexts, function(a, b=y) { a[1] == b[1] && a[2] == b[2] }, logical(1)))
condvars <- NULL
treatlevs <- NULL
reflevs <- NULL
if (!is.null(opt$comparisons)) {
selcomps <- comparisons[sel]
condvars <- vapply(selcomps, function(x){ x[[1]] }, character(1))
# stopifnot(length(unique(condvars)) == 1)
treatlevs <- vapply(selcomps, function(x){ as.character(unique(covars[[x[[1]]]])[x[[2]][1]]) }, character(1))
reflevs <- vapply(selcomps, function(x){ as.character(unique(covars[[x[[1]]]])[x[[2]][2]]) }, character(1))
if (!is.null(opt$fullFormula)) {
sanity <- vapply(condvars, function(x) { grepl(x, opt$fullFormula, fixed=TRUE) }, logical(1))
if (any(!sanity))
warning(paste(condvars[which(!sanity)], "not in formula", opt$fullFormula))
}
}
# Remove unused samples from the counts table,
# and ensure samples are in the same order as the covariates table.
subcovars <- covars[ covars[[contvar]] == contlev, ]
if (!is.null(opt$comparisons)) {
stopifnot(all( vapply(1:length(condvars), function(i) { reflevs[i] %in% subcovars[[condvars[i]]] }, logical(1)) ))
stopifnot(all( vapply(1:length(condvars), function(i) { treatlevs[i] %in% subcovars[[condvars[i]]] }, logical(1)) ))
}
cdn <- rownames(subcovars)
ctsn <- colnames(cts)
subcts <- cts[, ctsn[match(cdn, ctsn)]]
# Fill in non-detections
subcts[is.na(subcts)] <- 0
# DESeq2 begins here
designFormula <- ifelse(!is.null(opt$fullFormula),
opt$fullFormula,
paste0('~', paste(unique(condvars),collapse='+')) )
DDS <- DESeqDataSetFromMatrix(countData = round(subcts, digits=0),
colData = subcovars,
design = as.formula(designFormula) )
# To scale or not to scale
if (opt$prescaled) {
factors <- rep(1, times=length(cdn))
names(factors) <- cdn
sizeFactors(DDS) <- factors
}
# Prepare file name.
if (opt$prefix == '') {
# prefix <- paste(paste(contvar, contlev, sep='_'), designFormula, sep='.')
prefix <- paste(contvar, contlev, sep='_')
} else {
# prefix <- paste(opt$prefix, paste(contvar, contlev, sep='_'), designFormula, sep='.')
prefix <- paste(opt$prefix, paste(contvar, contlev, sep='_'), sep='.')
}
name <- paste(prefix, 'deseq2.html', sep='.')
#######################
#######################
subcomps <- NULL
if (!is.null(opt$comparisons)) {
subcomps <- lapply(1:length(condvars), function(i){ c(condvars[i], treatlevs[i], reflevs[i]) })
}
# Fire up an Rmd report
rmarkdown::render(opt$reportTemplate,
output_file = name,
output_dir = file.path(opt$baseDir, opt$resultsDir, opt$prefix),
intermediates_dir=file.path(opt$baseDir, opt$resultsDir, sub('deseq2.html', 'tmp', name)),
params=list(selfname = name,
ntop = opt$ntop,
pcutoff = opt$pcutoff,
lfcthreshold = opt$lfcthreshold,
baseDir = opt$baseDir,
resultsDir = file.path(opt$resultsDir, opt$prefix),
prefix = prefix,
dds = DDS,
minCount = opt$minCount,
minTPM = opt$minTPM,
filterBaseMean = opt$bmF,
reducedFormula = opt$reducedFormula,
longlabel=opt$label,
covars=covars,
widths = featLens,
specnorm=opt$specnorm,
rlog=!opt$vst,
comparisons=subcomps,
althyp=opt$altHypo,
fc4p=opt$fc4p)
)
}