Should I merge bacterial RNA-seq paired-end data?

Question

[biotech]

Hi,

I have bacterial RNA-seq data from paired-end reads (2x75). I'm interested in sense/antisense differential expression.

I would like to know if makes sense to merge R1 and R2 into one read (if overlaps) or maybe work only with 75 bp reads (R2 ones).

Thanks, Bernardo

[5heikki]

If your experiment was conducted so that the fragment size allowed pairs to overlap, merging them would make a lot of sense. So, finding out your fragment size would be the first thing..

[biotech]

200 bp but maybe shorter. The thing is that I need to convert paired-end to single-end data anyway in order to quantify antisense transcription too.

Answer 1

[Devon Ryan]

Why bother? Presuming that you used a stranded protocol when making the libraries, you can just align things as pairs and tell whichever program you use to do counting to observe strandedness of pairs. Both htseq-count and featureCounts can do that.

To get at the antisense transcription, one could just run the counters again and just tell them to use the other read as the strand reference (e.g., running htseq-count -s yes ... and then htseq-count -s reverse ...).

[biotech]

I did like you describe when working with single-end data. Is the same with paired-end data? I though that because one of the pairs aligns to reverse strand would count as antisense when is not. Also, I though longer reads would map better.

[Devon Ryan]

Any decent program that does counting will count pairs as a single unit, rather than individually, so yes it is the same.