csaw version: csaw_1.10.0

The call to combineTests below fails:

> tabcom <- combineTests(merged$id, results$table)
Error: length(ids) == nrow(tab) is not TRUE

> length(merged$id)
[1] 14051634
> nrow(results$table)
[1] 120695

When I get this error, what might the possible causes be?

My code is the following (and it works on many datasets, but fails on one particular for some reason):

data = readRDS(snakemake@input[["data"]])

design = read.table(snakemake@input[["design"]])
normfacs = unlist(read.table(snakemake@input[["normfacs"]]))
y = readRDS(snakemake@input[["estimate_disp"]])

library(csaw)
library(edgeR)
library(IRanges)

contrast = as.numeric(makeContrasts(snakemake@wildcards[["contrast"]], levels=colnames(design)))
fit <-  glmQLFit(y, design, contrast=contrast, robust=TRUE)
results <- glmQLFTest(fit, contrast=contrast)

# correct for multiple testing

print("correct for multiple testing")

merged <- mergeWindows(rowRanges(data), tol=1000L)
tabcom <- combineTests(merged$id, results$table)

print("writing output file")

df = as.data.frame(merged$region)
row.names = with(df, paste0(seqnames, "_", start, "_", end))

rownames(tabcom) = row.names

write.table(tabcom, snakemake@output[[1]])

Many thanks for your help!

If having the data available for running the code yourself might help debugging I can upload it to dropbox.

As the error message suggests, the lengths of the inputs don't match up. You have 14051634 elements in your rowRanges output but only 120695 rows in your results$table. This is most probably caused by the fact that you filtered on abundance at some point in your analysis, but then you kept the unfiltered object in in data. You need to recover the filtering vector and use it to subset data before applying combineTests, so that the genomic intervals in rowRanges correspond to the statistics in results$table.