phred scores and mapping

Question

[grant.hovhannisyan]

Do/How do Phred quality scores influence the read mapping? Can you share any software documentation describing that?

[h.mon]

I think most programs ignore read quality (but I don't have data to back my claim). There is an old thread (with links) stating BWA doesn't use base qualities ( does bwa mem use base qualities? ). The original information is here:

https://github.com/lh3/mem-paper/blob/master/response_to_aaron.md

[grant.hovhannisyan]

Thanks! A bit surprising to be honest, one would expect lower probability to map if you are not sure about the nucleotide.

[h.mon]

Bowtie2 seems to use read quality, although the manual is not entirely clear on how:

End-to-end alignment score example

A mismatched base at a high-quality position in the read receives a penalty of -6 by default. A length-2 read gap receives a penalty of -11 by default (-5 for the gap open, -3 for the first extension, -3 for the second extension). Thus, in end-to-end alignment mode, if the read is 50 bp long and it matches the reference exactly except for one mismatch at a high-quality position and one length-2 read gap, then the overall score is -(6 + 11) = -17.

The best possible alignment score in end-to-end mode is 0, which happens when there are no differences between the read and the reference.

Local alignment score example

A mismatched base at a high-quality position in the read receives a penalty of -6 by default. A length-2 read gap receives a penalty of -11 by default (-5 for the gap open, -3 for the first extension, -3 for the second extension). A base that matches receives a bonus of +2 be default. Thus, in local alignment mode, if the read is 50 bp long and it matches the reference exactly except for one mismatch at a high-quality position and one length-2 read gap, then the overall score equals the total bonus, 2 * 49, minus the total penalty, 6 + 11, = 81.

The best possible score in local mode equals the match bonus times the length of the read. This happens when there are no differences between the read and the reference.

[grant.hovhannisyan]

I am basically planning to simulate and map reads and was wondering whether the ability to simulate quality scores is important when choosing simulation software.