Hi everyone,

I was performing a comparison using DESeq2. Considering it is relevant for the case, I used a code I found here (DESeq2 baseMean values for each sample) to find the mean counts for each condition.

So, looking at a gene where both groups have zero counts, I obtained a significant p-value (quite small even, 5.7E-6). Since I had a warning about convergence, I also checked and it did converge.

Is this possible? I understand that the independentFiltering maintains the gene, since it has reads for other conditions but, considering the counts, it should be at least non-significant no?

The code is rather straightforward:

baseMeanPerLvl <- sapply( levels(dds_sub$CellType), function(lvl) 
      rowMeans(counts(dds_sub,normalized=TRUE)[,dds_sub$CellType == lvl, drop=F] ) )

resultsNames(dds_sub)
res1<-results(dds_sub,name =  "CellType_Neutrophils_vs_Eosinophil", alpha = 0.05)
res1<-lfcShrink(dds_sub,coef = "CellType_Neutrophils_vs_Eosinophil", type = "apeglm",res = res1)

Thanks in advance

Can you use contrast instead of name? This should resolve it.