FetureCounts low assigned rate and its weird total number counts?

Question

[Hughie]

Hi! myfriends,

I'm doing read count of my RNA-seq data using featurecounts, I found a pretty low assigned rate which ranges 20~30%, like below.

FeatureCounts result

I noticed that the total is not accord with my FastQC and STAR' report:

So, I used samtools flagstat to check:

I'm in doubt that whether featureCounts and samtools recounted multi-mapping reads? and if so, I wish your kindly suggestions, Thank you!

[GenoMax]

featureCounts does NOT count multi-mapped reads by default (which you don't want to do anyway). You can force it to do so using -M option. Check to see if you have rRNA contamination in your data if you see a large amount of multi-mapping.

[Kristoffer Vitting-Seerup]

I've previously seen examples were a lot of RNA-seq reads map to other parts of the genome (introns, intergenic) - could you test whether that is the case?

[Arindam Ghosh]

If " reads map to other parts of the genome (introns, intergenic)" should I remove the sample from downstream analysis?

Answer 1

[swbarnes2]

Those numbers don't look totally out of line to me. You have some reads aligning uniquely to a single feature, which featureCounts will count, some reads aligning uniquely, but not to features, and some reads aligning to multiple places.

I've run featureCounts on bams with secondary reads, and bams where I've filtered them away, and I get the same number of features (though the number of Unassigned_MultiMapping did change)