featureCounts

[SylvainL]

Hi all,

I am using (for the first time) featureCounts to get gene level counts in a RNAseq experiment... It is mouse RNAseq and polyA RNA were extracted...

A quick look at the results shows up to 15% of reads which are unassigned because of no overlap with any feature.

Do you normally have the same stats?

s.

[dpryan]

I just checked one dataset (mouse liver), where ~5% had no feature, ~3% were ambiguously mapped (a non-stranded protocol), and ~19% were not uniquely mapped. ~75% were counted if we exclude unmapped reads. This is similar in a mouse hippocampus dataset that I just checked. Obviously if you include unmapped reads then these percentages would go down a bit.

So, that unassigned percentage seems a bit high, but perhaps there's a good biological reason for that (e.g., in tumors or sperm).

[GenoMax]

Post #29. Response from Wei Shi (author of featureCounts): http://seqanswers.com/forums/showthr...t=49506&page=2

[SylvainL]

Hi dpryan,

I already removed the non uniquely mapped so yes, it's a bit lower if we consider all the mapped reads... and yes, it's sperm...

So I should not worry :9

Thanks a lot

[dpryan]

Yeah, sperm will inherently be different, since you have depletion of mRNAs. My sperm datasets aren't polyA enriched, they're ribo-depleted, so I don't have any numbers for comparison.