Can I treat pair-end reads as single end reads replicates?

[HSV-1]

I think mapping is no problem. Most of my pair-end reads were not mappable by Tophat. Only 15% of reads are mappble. By treating them as single end I could map more than 15%.
What is the potential problem?


Thanks!

[sdriscoll]

If you consider than each pair are just sequences from either end of single fragments then I think you can come to a conclusion about what you feel is OK to do. I personally think it's redundant to align both ends (if you're treating them as single end reads). I'd just align the left side reads since they are probably of higher quality than the right side.

[HSV-1]

If you consider than each pair are just sequences from either end of single fragments then I think you can come to a conclusion about what you feel is OK to do. I personally think it's redundant to align both ends (if you're treating them as single end reads). I'd just align the left side reads since they are probably of higher quality than the right side.

Thans for response. Good point.
Is left reads file normally suffixed with _1 and right ones with _2?

[sdriscoll]

in the reads i've gotten back from those HiSeq 2000 machines i get R1 and R2 for left and right. i'll bet they all use a 1 and 2 in some way to denote left and right.

[HSV-1]

If you consider than each pair are just sequences from either end of single fragments then I think you can come to a conclusion about what you feel is OK to do. I personally think it's redundant to align both ends (if you're treating them as single end reads). I'd just align the left side reads since they are probably of higher quality than the right side.

If so what I don't understand is what is the advantage of pair-end reads over single end reads? It seems to me that there is no much and significant difference between them considering each pair reads from a fragment that is not huge ~300-500bp, even smaller.

[westerman]

Paired end reads are useful in assembling large fragments larger than the insert size; e.g., usually for 1000+bp. PE reads are not as useful when your fragments are shorter than the insert size and you can often just toss the R2 reads away in these cases. It depends on your project.

That said, sequencing centers (I am a member of one) will sometimes throw what should be a SE project onto a sequencing run that has PE projects and run the SE project as PE. We figure if the person doesn't want the R2 reads they can ignore them.

[joachim.jacob]

Perhaps a strategy here:
- map the left reads alone
- look for a region of high coverage (e.g. chrom 21)
- extract the reads from that region
- extract the accompanying mate reads
- map the paired-end reads to that region again
- look at the difference in mapping between the two with a genomeviewer
- assemble stats of the two methods

Depending on your goal, this might be the most preferred and exhaustive method. Any comments and additions welcome!

While establishing gene expression values, my main concern is the detection of different splice variants with PE-reads. As nicely demonstrated in the cufflinks paper, a status-quo on the 'gene' level, may reveal difference on the transcript level ('isoform switching').

So if the genes that you know should change expression between condition (your control set), are not changing using SE reads mapping, it is worthwhile to do a PE read mapping, and look again to those genes, to investigate the isoforms if any.