How to Set Clipping Penalty in BWA ALN?

[svos]

Hi all!


I'm using BWA aln (v 0.7.10) to map short reads (75bp - 100bp) to hg19 genome. Is it possible to adjust the penalty for soft-clipping? I want to set it lower than mismatch penalty in order to prevent mismatches at read ends and to favor soft-clipping.

Alternatively, I could increase mismatch penalty, but what is the penalty for soft-clipping?


Thanks for your help!

[Brian Bushnell]

I'm not sure about bwa, but I do soft-clipping by trimming the ends to the point that the global score is maximized - therefore, the higher the mismatch penalty, the more soft-clipping will be done. Assuming that bwa operates in the same way, the amount of soft-clipping will be strictly dictated by the ratio of the match to mismatch (or gap, etc) scores. So to increase the amount of soft-clipping, just increase the mismatch penalty.

[svos]

At least for BWA mem algorithm this is not true. By default, it has a mismatch penalty of 4 and a clipping penalty of 5, so it rather introduces mismatches at the end of the reads instead of soft-slipping them.
In BWA mem, these penalties can be set by the user, but no such parameter exists for BWA aln...

[Brian Bushnell]

In that case, why not use bwa-mem? It's supposed to be better for reads over 70bp, anyway. Do you find that aln gives better results?

[svos]

Yes, I see mapping artifacts that I can not explain when I use BWA mem! BWA mem is set out to be used with reads having a length >= 100bp, but I'd still rather use BWA aln, as I get more reliable results (with both, 75bp and 100bp reads).