Dear there,

I know DESeq2 expects raw read counts as the inputs which is important for DESeq2's statistical model to hold (it has been emphasized many many times).

For singed-end reads, the counts should be the number of reads mapped to the gene. But, how about paired-end data, is it the number of mapped reads or fragments to the gene?

In another word, do I need to multiple 2 for the counts from htseq-count for PE data because htseq-count count reads pairs for PE data?

Thank you so much!

Best,

Xiaofei

No, use the counts of overlapping fragments. Each fragment is one observation of a gene, regardless of how many reads were produced from that one fragment.