Proper use of BWA-backtrack and BWA-MEM

Question

[cfarmeri]

Hi, Biostars.

I have a set of Illumina GA2 paired-end fastq files.

These reads of the fastqs has 76b length.

Before mapping, I trimmed barcode sequence from 5' end of both 1st and 2nd reads, finally, the reads have 63-67bp length.

BWA's author says following.

・reads < 70bp should be mapped by BWA-backtrack (BWA-backtrack can map 70bp< and <100bp reads, but they mention than these reads should be BWA-MEM algorithm.)

・reads > 100bp should be mapped by BWA-MEM. BWA-MEM algorithm is recommeded.

Any one has an idea which BWA-backtrack or BWA-MEM I should use?? Thanks!!!

Answer 1

[Brian Bushnell]

If the reads are < 70 bp after trimming, which it sounds like they are, then the official advice is to use BWA-backtrack. It does not matter how long they were initially.

Personally, if I were to use a BWA variant, I would choose BWA-mem over BWA-backtrack because it is so much faster and, in my testing, more accurate. But I have not tested it on < 70bp reads.

[cfarmeri]

Thanks, Brian.

I confused the 63-67bp is a value close to 70bp. First of all, I try BWA-backtrack algorithm.

Answer 2

[dariober]

Hi- Have a look also at this post When and why is bwa aln better then bwa mem?

Personally, these days I use bwa mem even when read length is < 70bp. I haven't tested it properly, but I find that for shorter reads, especially if paired-end, bwa mem becomes slower and more memory hungry than bwa aln.