Majority of reads counted in stranded=reverse using HTSeq?

[sdlmark]

Hello,

I'm new to working with RNAseq data and have a question about counting reads with HTSeq.

I have stranded data that was mapped using the default parameters in STAR. I am counting the reads using HTSeq with the parameters:
--type=exon
--mode=intersection-nonempty
--idattr=Parent
--format=bam
When I use the --stranded=yes option, fewer than 10% of reads are counted; the majority have "no feature." But when I use the --stranded=reverse option, over 80% of the reads are counted. Similarly, when I use the --stranded=no option, over 80% of the reads are counted.

I'm just wondering what could explain this pattern. Is this a typical outcome? Should I be concerned? I think that as a newbie I must be missing some key information about the stranded RNAseq protocol. Which, in this case, was the Illumina TruSeq Stranded protocol.

I'm pretty stumped so any insight into this matter would be greatly appreciated!

Thanks,

SM

[Michael.Ante]

Hi SM,

IMHO, you should not be worried. I would suggest, you get a bit more informed about your data (e.g. here https://www.biostars.org/p/64250/, the HTSeq-count manual ).
HTSeq-count is counting the reads, which align to the given exons. If you use the stranded option "yes", it checks whether the reads are in the same orientation as the transcript. Illumina's TruSeq Stranded protocol produces libraries, which are in reverse orientation to the transcripts' one.
The more overlapping genes you have, the stronger becomes the influence of the stranded-option in HTSeq-count.
The resulting number of usable reads depends on the pre-processing, the reads' mapping-quality, the annotation (e.g. RefSeq vs. Ensembl annotation), ....
Additionally, I'd recommend having a look at some genes, using e.g. the IGV browser and collect some statistics with RSeQC.

[sdlmark]

Thanks for the tips and info! I understand it better now.

SM

Hi SM,

IMHO, you should not be worried. I would suggest, you get a bit more informed about your data (e.g. here https://www.biostars.org/p/64250/, the HTSeq-count manual ).
HTSeq-count is counting the reads, which align to the given exons. If you use the stranded option "yes", it checks whether the reads are in the same orientation as the transcript. Illumina's TruSeq Stranded protocol produces libraries, which are in reverse orientation to the transcripts' one.
The more overlapping genes you have, the stronger becomes the influence of the stranded-option in HTSeq-count.
The resulting number of usable reads depends on the pre-processing, the reads' mapping-quality, the annotation (e.g. RefSeq vs. Ensembl annotation), ....
Additionally, I'd recommend having a look at some genes, using e.g. the IGV browser and collect some statistics with RSeQC.