htseq-count overlap modes

[is007]

Hi! I'm using htseq-count to count reads from my RNASeq experiment mapping to genes. I have an alignment file where all reads are mapped to the genome, and now I am just counting reads for each gene. It seems to work fine, but I'm trying to understand a specific detail about how it works. Basically, it has 3 overlap modes and I am using the 'union':



This is the documentation: http://www-huber.embl.de/users/ander...doc/count.html

I'm wondering if anyone knows to what extend does it count partially mapped reads (2nd row from the figure), i.e. by how much should the read overlap for it to be counted?

The figure implies it counts partial overlaps too, but does not specify what exactly the minimum overlap needs to be. I'm not finding this information in the documentation.

Thanks!

[dpryan]

A one base overlap counts as an overlap.

[is007]

Thanks, looks like you're right. I've come to the same conclusion just now inspecting their python source code. How's that meaningful?

I guess if there's a very small overlap (extreme case: just overlap 'A' from 'ATG' at beginning) then it's likely overlapping multiple genes, hence it's classified as 'ambiguous' and not counted?

[fanli]

Yes, but that depends on how dense the genes are in your genome, and how long your reads are

[dpryan]

Thanks, looks like you're right. I've come to the same conclusion just now inspecting their python source code. How's that meaningful?

There's no real meaning there other than how sets are dealt with.