-
Notifications
You must be signed in to change notification settings - Fork 0
/
differential_expression.R
502 lines (414 loc) · 21.2 KB
/
differential_expression.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
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
#Please refer to cacoa repository for full code
##' @title Validate parameters per cell type
##' @param raw.mats List of raw count matrices
##' @param cell.groups Named clustering/annotation factor with cell names
##' @param sample.groups Named factor with cell names indicating condition/sample, e.g., ctrl/disease
##' @param ref.level Reference cluster level in 'sample.groups', e.g., ctrl, healthy, wt
validateDEPerCellTypeParams <- function(raw.mats, cell.groups, sample.groups, ref.level) {
checkPackageInstalled("DESeq2", bioc=TRUE)
if (is.null(cell.groups)) stop('"cell.groups" must be specified')
if (is.null(sample.groups)) stop('"sample.groups" must be specified')
if (class(sample.groups) != "list") stop('"sample.groups" must be a list')
if (length(sample.groups) != 2) stop('"sample.groups" must be of length 2')
if (!all(unlist(lapply(sample.groups, function(x) class(x) == "character"))))
stop('"sample.groups" must be a list of character vectors')
if (!all(unlist(lapply(sample.groups, function(x) length(x) > 0))))
stop('"sample.groups" entries must be on length greater or equal to 1')
if (!all(unlist(lapply(sample.groups, function(x) all(x %in% names(raw.mats))))))
stop('"sample.groups" entries must be names of samples in the raw.mats')
if (is.null(ref.level)) stop('"ref.level" is not defined')
## todo: check samplegrousp are named
if (is.null(names(sample.groups))) stop('"sample.groups" must be named')
if (class(cell.groups) != "factor") stop('"cell.groups" must be a factor')
}
subsetMatricesWithCommonGenes <- function(cms, sample.groups=NULL) {
if (!is.null(sample.groups)) cms <- cms[unlist(sample.groups)]
common.genes <- do.call(intersect, lapply(cms, colnames))
cms %<>% lapply(function(m) m[, common.genes, drop=FALSE])
return(cms)
}
strpart <- function(x, split, n, fixed = FALSE) {
as.character(x) %>% strsplit(split, fixed=fixed) %>% sapply("[", n)
}
#' Add Z scores to DE results
#' @param df Data frame with the columns "pval", "padj" and "log2FoldChange"
#' @return Updated data frame with Z scores
#' @export
addZScores <- function(df) {
df$Z <- -qnorm(df$pval/2)
df$Z[is.na(df$Z)] <- 0
df$Za <- -qnorm(df$padj/2)
df$Za[is.na(df$Za)] <- 0
df$Z <- df$Z * sign(df$log2FoldChange)
df$Za <- df$Za * sign(df$log2FoldChange)
return(df)
}
#' @title Save DE results as JSON files
#' @param de.raw List of DE results
#' @param sample.groups Sample groups as named list, each element containing a vector of samples
#' @param saveprefix Prefix for created files (default=NULL)
#' @param dir.name Name for directory with results. If it doesn't exist, it will be created. To disable, set as NULL (default="JSON")
#' @param gene.metadata (default=NULL)
#' @param verbose Show progress (default=TRUE)
saveDEasJSON <- function(de.raw, saveprefix=NULL, dir.name="JSON", gene.metadata=NULL,
sample.groups=NULL, verbose=TRUE) {
if(!is.null(dir.name)) {
if(!dir.exists(dir.name)) dir.create(dir.name)
} else {
dir.name = "."
}
if (is.null(gene.metadata)) {
gene.metadata <- data.frame()
all.genes <- unique(unlist(lapply(de.raw, function(x) {
if (!is.null(x)) {
rownames(as.data.frame(x$res))
} else {
NULL
}
})))
gene.metadata <- data.frame(geneid=all.genes)
} else {
if(is.null(gene.metadata$gene.id)) stop("gene.metadata must contain $gene.id field")
}
lapply(sccore:::sn(de.raw %>% names()), function(ncc) {
if(verbose) print(ncc)
res.celltype <- de.raw[[ncc]]
res.table <- res.celltype$res %>% as.data.frame()
res.table$gene <- rownames(res.table)
res.table$significant <- res.table$padj < 0.05
res.table$log2FoldChange[is.na(res.table$log2FoldChange)] <- 0
res.table$rowid <- 1:nrow(res.table)
all.genes <- rownames(res.table)
cm <- res.celltype$cm
ilev <- lapply(sample.groups, function(sg) {
sg <- sg[sg %in% colnames(cm)]
cm.tmp <- cm[,sg]
cm.tmp <- as.matrix(cm.tmp)
rownames(cm.tmp) <- rownames(cm)
## calculate cpm
cpm <- sweep(cm.tmp, 2, apply(cm.tmp,2, sum), FUN='/')
cpm <- log10(cpm * 1e6 + 1)
snames1 <- colnames(cpm)
## Put genes in order
cpm <- cpm[all.genes,]
colnames(cpm) <- NULL;
rownames(cpm) <- NULL;
list(snames=snames1, val=as.matrix(cpm))
})
snames <- names(res.celltype$sample.groups)
## convert to json
tojson <- list(
res = res.table,
genes = all.genes,
ilev = ilev,
snames = snames)
y <- jsonlite::toJSON(tojson)
file <- paste0(dir.name, "/", saveprefix, make.names(ncc), ".json")
write(y, file)
NULL
})
toc.file <- paste0(dir.name, "/toc.html")
s <- c(list('<html><head><style>
table {
font-family: arial, sans-serif;
border-collapse: collapse;
width: 100%;
}
td, th {
border: 1px solid #dddddd;
text-align: left;
padding: 8px;
}
tr:nth-child(even) {
background-color: #dddddd;
}
</style></head><body><table>'),
lapply(names(de.raw), function(n)
paste0('<tr><td><a href="deview.2.html?d=', saveprefix, make.names(n),'.json">', n, '</a></td></tr>')),
list('</table></body></html>')
) %>% paste(collapse='\n')
write(s, file=toc.file)
}
prepareSamplesForDE <- function(sample.groups, resampling.method=c('loo', 'bootstrap', 'fix.cells', 'fix.samples'),
n.resamplings=30, n.biosamples=NULL) {
resampling.method <- match.arg(resampling.method)
if (resampling.method == 'loo') {
samples <- unlist(sample.groups) %>% sn() %>% lapply(function(n) lapply(sample.groups, setdiff, n))
} else if (resampling.method == 'bootstrap') {
# TODO: Do we ever use bootstrap? It seems that including the same sample many times
# reduces variation and skews the analysis
samples <- (1:n.resamplings) %>% setNames(paste0('bootstrap.', .)) %>%
lapply(function(i) lapply(sample.groups, function(x) sample(x, length(x), replace=TRUE)))
} else { # 'fix.cells' or 'fix.samples'
samples <- (1:n.resamplings) %>% setNames(., paste0('fix.', .)) %>% lapply(function(i) sample.groups)
}
return(samples)
}
#' Differential expression using different methods (deseq2, edgeR, wilcoxon, ttest) with various covariates
#' @param raw.mats list of counts matrices; column for gene and row for cell
#' @param cell.groups factor specifying cell types (default=NULL)
#' @param sample.groups a list of two character vector specifying the app groups to compare (default=NULL)
#' @param ref.level Reference level in 'sample.groups', e.g., ctrl, healthy, wt (default=NULL)
#' @param common.genes Only investigate common genes across cell groups (default=FALSE)
#' @param cooks.cutoff cooksCutoff for DESeq2 (default=FALSE)
#' @param min.cell.count (default=10)
#' @param independent.filtering independentFiltering for DESeq2 (default=FALSE)
#' @param n.cores Number of cores (default=1)
#' @param return.matrix Return merged matrix of results (default=TRUE)
#' @param covariates list of covariates to include; for example, cdr, sex or age
#' @param meta.info dataframe with possible covariates; for example, sex or age
#' @param test DE method: deseq2, edgeR, wilcoxon, ttest
#' @export
estimateDEPerCellTypeInner=function(raw.mats, cell.groups=NULL, s.groups=NULL, ref.level=NULL, target.level=NULL,
common.genes=FALSE, cooks.cutoff=FALSE, min.cell.count=10, max.cell.count=Inf,
independent.filtering=TRUE, n.cores=4, return.matrix=TRUE, fix.n.samples=NULL,
verbose=TRUE, test='Wald', meta.info=NULL, gene.filter=NULL) {
# Validate input
validateDEPerCellTypeParams(raw.mats, cell.groups, s.groups, ref.level)
tmp <- tolower(strsplit(test, split='\\.')[[1]])
test <- tmp[1]
test.type <- ifelse(is.na(tmp[2]), '', tmp[2])
# Filter data and convert to the right format
if (verbose) message("Preparing matrices for DE")
if (common.genes) {
raw.mats %<>% subsetMatricesWithCommonGenes(s.groups)
} else {
gene.union <- lapply(raw.mats, colnames) %>% Reduce(union, .)
raw.mats %<>% plapply(sccore:::extendMatrix, gene.union, n.cores=n.cores, progress=verbose, mc.preschedule=TRUE)
}
cm.bulk.per.samp <- raw.mats[unlist(s.groups)] %>% # Only consider samples in s.groups
lapply(collapseCellsByType, groups=cell.groups, min.cell.count=min.cell.count, max.cell.count=max.cell.count) %>%
.[sapply(., nrow) > 0] # Remove empty samples due to min.cell.count
cm.bulk.per.type <- levels(cell.groups) %>% sn() %>% lapply(function(cg) {
tcms <- cm.bulk.per.samp %>%
lapply(function(cm) if (cg %in% rownames(cm)) cm[cg, , drop=FALSE] else NULL) %>%
.[!sapply(., is.null)]
if (length(tcms) == 0) return(NULL)
tcms %>% {set_rownames(do.call(rbind, .), names(.))} %>% `mode<-`('integer') %>%
.[,colSums(.) > 0,drop=FALSE]
}) %>% .[sapply(., length) > 0] %>% lapply(t)
## Adjust s.groups
passed.samples <- names(cm.bulk.per.samp)
if (verbose && (length(passed.samples) != length(unlist(s.groups))))
warning("Excluded ", length(unlist(s.groups)) - length(passed.samples), " sample(s) due to 'min.cell.count'.")
s.groups %<>% lapply(intersect, passed.samples)
# For every cell type get differential expression results
if (verbose) message("Estimating DE per cell type")
de.res <- names(cm.bulk.per.type) %>% sn()%>% plapply(function(l) {
tryCatch({
cm <- cm.bulk.per.type[[l]]
if (!is.null(gene.filter)) {
gene.to.remain <- gene.filter %>% {rownames(.)[.[,l]]} %>% intersect(rownames(cm))
cm <- cm[gene.to.remain,,drop=FALSE]
}
cur.s.groups <- lapply(s.groups, intersect, colnames(cm))
if (!is.null(fix.n.samples)) {
if (min(sapply(s.groups, length)) < fix.n.samples) stop("The cluster does not have enough samples")
cur.s.groups %<>% lapply(sample, fix.n.samples)
cm <- cm[, unlist(cur.s.groups), drop=FALSE]
}
## Generate metadata
meta.groups <- colnames(cm) %>% lapply(function(y) {
names(cur.s.groups)[sapply(cur.s.groups, function(x) any(x %in% y))]
}) %>% unlist() %>% as.factor()
if (length(levels(meta.groups)) < 2) stop("The cluster is not present in both conditions")
if (!ref.level %in% levels(meta.groups)) stop("The reference level is absent in this comparison")
meta <- data.frame(sample.id=colnames(cm), group=relevel(meta.groups, ref=ref.level))
## External covariates
if (is.null(meta.info)) {
design.formula <- as.formula('~ group')
} else {
design.formula <- c(colnames(meta.info), 'group') %>%
paste(collapse=' + ') %>% {paste('~', .)} %>% as.formula()
meta %<>% cbind(meta.info[meta$sample.id, , drop=FALSE])
}
if (test %in% c('wilcoxon', 't-test')) {
cm <- normalizePseudoBulkMatrix(cm, meta=meta, design.formula=design.formula, type=test.type)
res <- estimateDEForTypePairwiseStat(cm, meta=meta, target.level=target.level, test=test)
} else if (test == 'deseq2') {
res <- estimateDEForTypeDESeq(
cm, meta=meta, design.formula=design.formula, ref.level=ref.level, target.level=target.level,
test.type=test.type, cooksCutoff=cooks.cutoff, independentFiltering=independent.filtering
)
} else if (test == 'edger') {
res <- estimateDEForTypeEdgeR(cm, meta=meta, design.formula=design.formula)
} else if (test == 'limma-voom') {
res <- estimateDEForTypeLimma(cm, meta=meta, design.formula=design.formula, target.level=target.level)
}
if (!is.na(res[[1]][1])) {
res <- addZScores(res) %>% .[order(.$pvalue, decreasing=FALSE),]
}
if (return.matrix)
return(list(res = res, cm = cm, meta=meta))
return(res)
}, error = function(err) NA)
}, n.cores=n.cores, progress=verbose) %>% .[!sapply(., is.logical)]
if (verbose) {
dif <- setdiff(levels(cell.groups), names(de.res))
if (length(dif) > 0) {
message("DEs not calculated for ", length(dif), " cell group(s): ", paste(dif, collapse=', '))
}
}
return(de.res)
}
normalizePseudoBulkMatrix <- function(cm, meta=NULL, design.formula=NULL, type='totcount') {
if (type == 'deseq2') {
cnts.norm <- DESeq2::DESeqDataSetFromMatrix(cm, meta, design=design.formula) %>%
DESeq2::estimateSizeFactors() %>% DESeq2::counts(normalized=TRUE)
} else if (type == 'edger') {
cnts.norm <- edgeR::DGEList(counts=cm) %>% edgeR::calcNormFactors() %>% edgeR::cpm()
} else if (type == 'totcount') {
# the default should be normalization by the number of molecules!
cnts.norm <- prop.table(cm, 2) # Should it be multiplied by median(colSums(cm)) ?
}
return(cnts.norm)
}
estimateDEForTypePairwiseStat <- function(cm.norm, meta, target.level, test) {
if (test == 'wilcoxon') {
res <- scran::pairwiseWilcox(cm.norm, groups = meta$group)$statistics[[1]] %>%
data.frame() %>% setNames(c("AUC", "pvalue", "padj"))
} else if (test == 't-test') {
res <- scran::pairwiseTTests(cm.norm, groups = meta$group)$statistics[[1]] %>%
data.frame() %>% setNames(c("AUC", "pvalue", "padj"))
}
# TODO: log2(x + 1) does not work for total-count normalization
res$log2FoldChange <- log2(cm.norm + 1) %>% apply(1, function(x) {
mean(x[meta$group == target.level]) - mean(x[meta$group != target.level])})
return(res)
}
estimateDEForTypeDESeq <- function(cm, meta, design.formula, ref.level, target.level, test.type, ...) {
res <- DESeq2::DESeqDataSetFromMatrix(cm, meta, design=design.formula)
if (test.type == 'wald') {
res %<>% DESeq2::DESeq(quiet=TRUE, test='Wald')
} else {
res %<>% DESeq2::DESeq(quiet=TRUE, test='LRT', reduced = ~ 1)
}
res %<>% DESeq2::results(contrast=c('group', target.level, ref.level), ...) %>% as.data.frame()
res$padj[is.na(res$padj)] <- 1
return(res)
}
estimateDEForTypeEdgeR <- function(cm, meta, design.formula) {
design <- model.matrix(design.formula, meta)
qlf <- edgeR::DGEList(cm, group = meta$group) %>%
edgeR::calcNormFactors() %>%
edgeR::estimateDisp(design = design) %>%
edgeR::glmQLFit(design = design) %>%
edgeR::glmQLFTest(coef=ncol(design))
res <- qlf$table %>% .[order(.$PValue),] %>% set_colnames(c("log2FoldChange", "logCPM", "stat", "pvalue"))
res$padj <- p.adjust(res$pvalue, method="BH")
return(res)
}
estimateDEForTypeLimma <- function(cm, meta, design.formula, target.level) {
mm <- model.matrix(design.formula, meta)
fit <- limma::voom(cm, mm, plot = FALSE) %>% limma::lmFit(mm)
contr <- paste0('group', target.level) %>% limma::makeContrasts(levels=colnames(coef(fit)))
res <- limma::contrasts.fit(fit, contr) %>% limma::eBayes() %>% limma::topTable(sort.by="P", n=Inf) %>%
set_colnames(c('log2FoldChange', 'AveExpr', 'stat', 'pvalue', 'padj', 'B'))
return(res)
}
#' Summarize DE Resampling Results
#' @param var.to.sort Variable to calculate ranks
summarizeDEResamplingResults <- function(de.list, var.to.sort='pvalue') {
de.res <- de.list[[1]]
for (cell.type in names(de.res)) {
genes.init <- genes.common <- rownames(de.res[[cell.type]]$res)
mx.stat <- matrix(nrow = length(genes.common), ncol = 0, dimnames = list(genes.common,c()))
for (i in 2:length(de.list)) {
if (!(cell.type %in% names(de.list[[i]]))) next
genes.common <- intersect(genes.common, rownames(de.list[[i]][[cell.type]]))
mx.stat <- cbind(mx.stat[genes.common,,drop=FALSE],
de.list[[i]][[cell.type]][genes.common, var.to.sort,drop=FALSE])
}
if (ncol(mx.stat) == 0) {
warning("Cell type ", cell.type, " was not present in any subsamples")
next
}
mx.stat <- apply(mx.stat, 2, rank)
stab.mean.rank <- rowMeans(mx.stat) # stab - for stability
stab.median.rank <- apply(mx.stat, 1, median)
stab.var.rank <- apply(mx.stat, 1, var)
de.res[[cell.type]]$res$stab.median.rank <- stab.median.rank[genes.init]
de.res[[cell.type]]$res$stab.mean.rank <- stab.mean.rank[genes.init]
de.res[[cell.type]]$res$stab.var.rank <- stab.var.rank[genes.init]
# Save subsamples
de.res[[cell.type]]$subsamples <- lapply(de.list[2:length(de.list)], `[[`, cell.type)
}
return(de.res)
}
appendStatisticsToDE <- function(de.list, expr.frac.per.type) {
for (n in names(de.list)) {
de.list[[n]]$res %<>% mutate(Gene=rownames(.), CellFrac=expr.frac.per.type[Gene, n],
SampleFrac=Matrix::rowMeans(de.list[[n]]$cm > 0)[Gene]) %>%
as.data.frame(stringsAsFactors=FALSE) %>% set_rownames(.$Gene)
}
return(de.list)
}
getExpressionFractionPerGroup <- function(cm, cell.groups) {
cm@x <- as.numeric(cm@x > 1e-10)
fracs <- collapseCellsByType(cm, cell.groups, min.cell.count=0) %>%
{. / as.vector(table(cell.groups)[rownames(.)])} %>% Matrix::t()
return(fracs)
}
estimateDEPerCellType=function(cell.groups=self$cell.groups, sample.groups=self$sample.groups,
ref.level=self$ref.level, target.level=self$target.level, name='de',
test='DESeq2.Wald', resampling.method=NULL, n.resamplings=30, seed.resampling=239,
min.cell.frac=0.05, covariates=NULL, common.genes=FALSE, n.cores=self$n.cores,
cooks.cutoff=FALSE, independent.filtering=FALSE, min.cell.count=10,
n.cells.subsample=NULL, verbose=self$verbose, fix.n.samples=NULL, ...) {
set.seed(seed.resampling)
if (!is.list(sample.groups)) {
sample.groups %<>% {split(names(.), . == ref.level)} %>% setNames(c(target.level, ref.level))
}
possible.tests <- c('DESeq2.Wald', 'DESeq2.LRT', 'edgeR',
'Wilcoxon.edgeR', 'Wilcoxon.DESeq2', 'Wilcoxon.totcount',
't-test.edgeR', 't-test.DESeq2', 't-test.totcount',
'limma-voom')
if (tolower(test) == tolower('DESeq2')) test <- paste(test, 'Wald', sep='.')
if (tolower(test) %in% tolower(c('Wilcoxon', 't-test'))) test <- paste(test, 'edgeR', sep='.')
if (!(tolower(test) %in% tolower(possible.tests)))
stop('Test ', test, ' is not supported. Available tests: ', paste(possible.tests, collapse=', '))
# s.groups.new contains list of case/control groups of samples to run DE on.
# First element in s.groups.new corresponds to the initial grouping.
if (!is.null(n.cells.subsample) && is.null(resampling.method)) resampling.method <- 'fix.cells'
s.groups.new <- list(initial=sample.groups)
max.cell.count <- Inf
fix.samples <- NULL
# If resampling is defined, new contrasts will append to s.groups.new
if (!is.null(resampling.method) && (n.resamplings != 0)) {
s.groups.new %<>% c(
prepareSamplesForDE(sample.groups, resampling.method=resampling.method, n.resamplings=n.resamplings)
)
if (resampling.method == 'fix.samples') {
if (is.null(fix.n.samples)) stop("fix.n.samples must be provided for resampling.method='fix.samples'")
fix.samples <- fix.n.samples
}
}
if (!is.null(n.cells.subsample)) {
if (verbose) message('Number of cell counts is fixed to ', n.cells.subsample)
max.cell.count <- min.cell.count <- n.cells.subsample
}
raw.mats <- extractRawCountMatrices(self$data.object, transposed=TRUE)
expr.fracs <- self$getJointCountMatrix() %>% getExpressionFractionPerGroup(cell.groups)
gene.filter <- (expr.fracs > min.cell.frac)
# parallelize the outer loop if subsampling is on
outer.multicore <- (length(s.groups.new) >= n.cores) && (n.cores > 1)
inner.verbose <- (length(s.groups.new) == 1) || (!outer.multicore && verbose > 1)
de.res <- names(s.groups.new) %>% sn() %>% plapply(function(resampling.name) {
estimateDEPerCellTypeInner(
raw.mats=raw.mats, cell.groups=cell.groups, s.groups=s.groups.new[[resampling.name]],
ref.level=ref.level, target.level=target.level, common.genes=common.genes,
cooks.cutoff=cooks.cutoff, min.cell.count=min.cell.count, max.cell.count=max.cell.count,
independent.filtering=independent.filtering, test=test, meta.info=covariates, gene.filter=gene.filter,
fix.n.samples=(if (resampling.name == 'initial') NULL else fix.samples),
n.cores=ifelse(outer.multicore, 1, n.cores),
return.matrix=(resampling.name == 'initial'),
verbose=(inner.verbose & verbose), ...
)
}, n.cores=ifelse(outer.multicore, n.cores, 1), progress=(!inner.verbose & verbose))
# if resampling: calculate median and variance on ranks after resampling
de.res <- if (length(de.res) > 1) summarizeDEResamplingResults(de.res) else de.res[[1]]
de.res %<>% appendStatisticsToDE(expr.fracs)
self$test.results[[name]] <- de.res
# TODO: add overall p-adjustment
return(invisible(self$test.results[[name]]))
}