Does featureCounts count/include the unmapped reads?

Question

[Pahi]

Hi!

Recently I had a few samples, where I did a STAR alignment, and there were a lot of unmapped reads.

I ran featureCounts, and got my results, and also did all the downstream analysis (DE, GO, etc.).

I was wondering, if featureCounts also include the unmapped reads? Does the unmapped reads effect the normalization in the downstream process?

Can you please help me in understanding this problem? Is there an article about this? Or I'm just overthinking and the unmapped reads doesn't matter at all.

Thank you for your help!

Answer 1

[Michael]

Or I'm just overthinking and the unmapped reads doesn't matter at all.

I think, in general, unmapped reads are mostly ignored in the literature. At least, unmapped reads are nothing that can be fixed at the featureCounts step. If at all, it needs to be addressed during mapping by taking into account possible reasons. These are for example contamination, sequencing errors, gaps in the reference, and also that many RNAseq samples contain multiple organisms' RNA, not just the one we are interested in. In most cases, these reads cannot be cured and it should be safe to ignore them for practical purposes. If the mapping rate is very low, it is wise to check the reasons for this.

[Pahi]

Thank you for your answer! So far this is what I'm thinking also. As for the mapping rate, we know what is the reason for the poor quality, but I'm doing the analysis for another group, who don't want to redo the sequencing, so I have to go with the poor mapping! :)