RNA-Seq: Merge files before or after counting?

[bodhisattvax]

Hi all

We've got a few RNA-Seq experiments planned running multiplexed samples, such that all samples to be analysed are run in a single library. Because of the high number of samples involved (30-60), we are running several lanes to make up the sequencing depth.

So prior to carrying out differential gene expression analyses, I would have to merge data for the same sample across the different lanes (e.g. for Sample A in lanes 1,2 and 3: A1 + A2 + A3).

When would be the appropriate time to merge the data?
Should I merge the aligned bams of A1,A2,A3 and then count reads in genes e.g. with HTSeq-count?
Or should I count A1, A2 and A3 separately then add up the counts for all the genes across lanes?

Does anyone have any experience or opinions on this?

[Simon Anders]

Should give the same results, shouldn't it?

The advantage of counting first is that you can check for lane effects, though these are so rare that I usually don't bother, the advantage of merging first is that you have less files to keep track of.

[bodhisattvax]

Thanks Simon

Yes, that's what I'd thought - just wanted to check if I was missing something obvious ..

Cheers