what happens with read duplicates?

[splaisan]

I routinely mark (I do not delete) my duplicate reads using Picard MarkDuplicate wether I do gDNA seq or RNA Seq but I still ignore (besides for GATK) which downstream software do take the marking into account and HOW.

Could someone comment on the way one should handle duplicate reads (optical & PCR types) in counting for RNASeq or variant analysis.

If you know that a particular piece of software does something by default, please let me know!

HTSeq
featureCount (has a arg --ignoreDup; who uses it?)
bcftools/samtools
varscan

Thanks for your info

[dpryan]

It's unwise to ignore duplicates (at least those produced by picard) in RNAseq, it'll just deflate correct counts for highly expressed genes. One could devise a more RNAseq-centric duplicate marker that took overall gene coverage into account when doing the marking, but I suspect that the gain from doing that is minimal (after all, you should already be using sufficient replicates to allow things like filtering by Cook's distance to work).

[splaisan]

Thanks Ryan but I did not ask if duplicates should be removed which I know they should not be.

[dpryan]

I know, which is why I didn't mention anything about removing them.