Quality trimming with bwa-mem

[shmibble]

Hi,
When using bwa-mem, is it necessary to use a different program for quality trimming prior to alignment? Since FASTX toolkit doesn't work with paired end reads I was thinking trimmomatic perhaps. But is this even necessary since bwa-mem has the -q flag for quality trimming?
And if I should use trimmomatic, how do I align the output with bwa-mem? Because as I understand it creates outputs for mapped reads and unmapped reads seperately.
Hope my question makes sense. I'm kinda new at this.

Thanks!

[dariober]

For adapter and quality trimming in paired-end reads you can use cutadapt 1.8+:

For example:

cutadapt -a AGATCGGAAGAGC -A AGATCGGAAGAGC -q 15 -o fq1.trim.fq -p fq2.trim.fq fq1.fq fq2.fq

However, it's seems with bwa mem is not necessary to trim for quality. In fact it's not necessary to trim adapters either. I think the GATK guidelines do not recommend adapter & quality trimming.

[shmibble]

Thanks for the reply. So You're saying bwa mem should work better without trimming at all before using it?

[dariober]

I haven't done any benchmark. Personally I do trim adapters and recently I do not quality trim. But yes, bwa mem should handle raw reads with adapters and bad quality ends "correctly" (...for some definition of correct!). Try aligning a bunch of reads with and without trimming and see what happens...

[GenoMax]

NGS aligners will work without trimming. As to whether they will work "better" that is hard to say. Most aligners will "soft"-clip reads (parts that don't align). That said you want to start with minimal (ideally none) extraneous sequence that does not belong to your sample when you start the analysis.

Trimmomatic/BBDuk.sh from BBMap suite are additional examples of trimming programs that will accept paired-end files and maintain the read order post trimming.

[shmibble]

Great. Will do. Thanks!

[Brian Bushnell]

I would always recommend adapter-trimming; adapter sequence at the ends of reads will never help you in mapping, and can only hurt. Quality-trimming is much more case-by-case, depending on your data quality, intended use, and specific software. I don't generally recommend it before mapping unless you have sufficiently low-quality data that it is reducing your mapping rate. Even low-quality bases can increase the ability to uniquely place a read, or to distinguish between substitutions and indels.