View Single Post
Old 02-26-2014, 04:30 PM   #6
Location: Australia, Brisbane

Join Date: Nov 2012
Posts: 13

Hi Simon,

I have had a look at the alignments.
When the counts between stranded and unstraded librarires agree reasonably well, it looks alright. I can see that when they differ strongly it is because the locus is particulary complex, with a lot of antisense transcription and overlapping genes, which are pickep up only by the stranded libraries.

I've also noticed that in a lot of cases, genes that are well supported by alignment reads, have a count of 0. This happens more often for unstranded libraries, I guess particularly when there is antisense transcription. But it also happens for stranded libraries. Highly expressed genes with good alignment support have a count 0.

I wonder if running htseq-count with
-m union
could have influenced this.
For each developmental stage, around 15-20 million counts are ambigous.

Would -m intersection_nonempty more appropriate in your opinion?

Thanks for help
uqfgaiti is offline   Reply With Quote