Hi All: I am using EdgeR to analyze the differential gene expression but today I met a problem I never seen before. In the step of filtering out all the low expression gene using filterByExpr(x, group = group), it had all my counts filtered out.

My code is following:

files <- c("E11Rep1.txt", "E11Rep2.txt", "E14Rep1.txt", "E14Rep2.txt", "E18Rep1.txt", "E18Rep2.txt", "Adult_Rep1.txt", "Adult_Rep2.txt")
read.delim(files[1], nrows = 5)
group <- as.factor(c("E11Rep1", "E11Rep2", "E14Rep1", "E14Rep2", "E18Rep1", "E18Rep2", "AdultRep1", "AdultRep2"))
x$samples$group <- group
lane <- as.factor(c("Rep1", "Rep2", "Rep1", "Rep2","Rep1", "Rep2", "Rep1", "Rep2") )
x$samples$lane <- lane
library(Mus.musculus)
geneid <- row.names(x)
genes <- select(Mus.musculus, keys = geneid, columns = c("SYMBOL", "TXCHROM"), keytype = "ENTREZID")
genes <- genes[!duplicated(genes$ENTREZID), ]
x$genes <- genes
cpm <- cpm(x)
lcpm <- cpm(x, log=TRUE)
L <- mean(x$samples$lib.size)*1e-6
M <- median(x$samples$lib.size)*1e-6
c(L, M)
table(rowSums(x$counts==0)==8)
keep.exprs <- filterByExpr(x, group = group)

The output is:

> Warning message: In min(n[n > 1L]) : no non-missing arguments to min;
> returning Inf

I have used x$count to check the count number and they are not that low:

    head(x$counts)

> Samples Tags        E11Rep1 E11Rep2 E14Rep1 E14Rep2 E18Rep1 E18Rep2
> Adult_Rep1 Adult_Rep2   497097        199     190     617     889   
> 1148    1761        961        884   100503874      82      97     234
> 281     395     504        607        527   100038431       0       0 
> 1       8      18      25         19         21   19888           6   
> 19       9      17       6       9         13          1   20671      
> 1704    1334    1544    2321    1412    2160       1657       1539  
> 27395        7984    5541    3778    5846    3126    5332       1736  
> 1618

can any one tell me why this happened?

Thank you very much!

There are two problems:

1. group should contain group labels, not sample labels.

2. You are using out-of-date versions of edgeR and Bioconductor. If you don't want to install Bioconductor 3.10, then at least reinstall edgeR for the Bioconductor release you are using.

Correcting either or both of these problems will fix the filtering.