A feature request for short read aligners

[ymc]

I find that the major short read aligners map poorly for highly polymorphic genes like HLA or Cytochrome P450. This is expected because for alleles that differ a lot from the reference allelle, there will be too many mismatches.

I think we can solve this problem by reading a dbsnp vcf that contains possible alleles and allele frequencies during alignment, then treat calls differ from the reference to be matches instead of mismatches if they exceed a certain allele frequency, e.g. 1%. I think this feature can improve the percentage of reads mapped greatly.

Can any short read aligner authors add this feature? Or is it already available in some aligners?

Thanks!

[kopi-o]

It can be done in http://seqanswers.com/wiki/GSNAP and probably some other aligners.

[ymc]

It can be done in http://seqanswers.com/wiki/GSNAP and probably some other aligners.

Thanks for pointing it out. I will give it a try.

Do you know if GSNAP gives higher mapping percentage than bwa?

[sphil]

I find that the major short read aligners map poorly for highly polymorphic genes like HLA or Cytochrome P450. This is expected because for alleles that differ a lot from the reference allelle, there will be too many mismatches.

you can set up the mismatches in every aligner to a number of your choice but keep in mind that highering the acceptable number of mismatches will raise runtime more or less dramatically!

However, I like the idea adding some DB-information to the mapping algorithms!