Question regarding almost identical features in exon_grouped_counts output

[wufishy]

Hi,

I am currently utilising the exon_grouped_counts output from IsoQuant to perform some differential exon usage calculation.
I found that some exons that have the exact same chr, start, end, strand, gene, and group_id will be assigned different flags and have different read counts, for example:

chr	start	end	strand	gene_ids	group_id	include_counts	flags
chr1	6197652	6197756	-	ENSG00000116251.12	group 1	15	IC
chr1	6197652	6197756	-	ENSG00000116251.12	group 1	4223	ISC

I am wondering why are these exons classified as different features. Is it because they are the result of distinct intron graph paths (that do not converge over this exon) or another reason? Am I right to conclude that they are safe to sum together if all I want to know in this instance is the exon usage.

Thank you!
Best wishes,
Jack

[wufishy]

Hi, to add to this

If you set the option --transcript_quantification unique_only, will non-unique mappings contribute to the exon counts?

Best wishes,
Jack

[andrewprzh]

Dear @wufishy

Sorry for the delayed response. This is odd, which annotation did you use? I would like to reproduce the problem on my side.

If you set the option --transcript_quantification unique_only, will non-unique mappings contribute to the exon counts?

No, this affects only transcript quantification. In fact, using only uniquely assigned transcripts can reduce coverage and statistical power for subsequent exon testing. We consider that a read covering an annotated exon serves as sufficient evidence regardless of whether a read is uniquely assigned to an isoform.

Best
Andrey

[andrewprzh]

I reproduced the bug and will fix it soon.
So far it's safe just to sum the counts for both lines.

[wufishy]

Thank you! All the best.

[andrewprzh]

@wufishy

Thanks a lot for the report! This bug affected only small portion of genes and was not easy to spot.
Now fixed in https://github.com/ablab/IsoQuant/releases/tag/v3.13.1

Best
Andrey