Overlapping 5UTR-CDS features with HTSeq: Ambiguous?

[Rivalyn]

My colleagues and I are interested in the distribution of reads that fall within the 5'UTR, CDS and 3'UTR of a given transcript. We made a feature count table with HTSeq using the --type=<feature type> argument. The invocation looks like:

python -m HTSeq.scripts.count --format bam --order pos --stranded yes --type CDS file.bam pathto/ucsc_refseq.gtf > CDS_readcount.txt

We did the same for 5'UTR and 3'UTR.

My question is, what happens when a read spans both the 5'UTR and the CDS? My understanding of the manual is that the read will be counted as ambiguous in union mode and as 5'UTR in intersection-strict or intersection-nonempty mode, but it's important for our downstream analysis to be certain of this.

[dpryan]

With "--type CDS" you're ignoring UTRs completely, so if the alignment overlaps a UTR and a CDS it'll get counted unless you use "intersection-strict" mode. This is similar for any other "--type".

[Rivalyn]

Ok, so then it will get counted twice? Once in the UTR output and once in the CDS output?

[dpryan]

Yes, such alignments will be included in both counts files.

[Rivalyn]

Thanks so much!