Aligning Paired-End Reads In Single-End Mode

Question

[GPR]

Hello,

I have a question on how to align paired-end reads.

In cases of very large fastq files, which make aligners like TopHat crash in a server with limited memory in RAM, I have seen people align one pair mate at a time, to prevent TopHat from crashing. I cann't come up with an argument against doing this, but am tempted to reason that one should always align paired-end reads in "paired-end mode".

My question is: How are downstream analyses affected if one aligns paired-end reads one pair mate at a time, setting BWA or Bowtie/TopHat to single-end mode?

Thanks, G.

Answer 1

[Farhat]

Bowtie and BWA both index the genome and not the reads, so the RAM requirements are proportional to the genome size, and not the size of the FASTQ files. That aside, using single-end mode for paired end reads will give you more non-unique alignments since in case one of the pairs falls in a repeat region but the other doesn't, you would still be able to align it properly using information from the mate.

[GPR]

Comment appreciated, thanks.