alexdobin / STAR

RNA-seq aligner

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Any way to increase penalty for a mismatch?

karlestira opened this issue · comments

I use STAR for FUSION detect. But some times the fusion partner's sequence is similar to the next exon(in gtf file). Mapping to fusion partner lead to 0 mismatches, and mapping to next exon lead to 5 mismatches in 16bp(this is very low quality).

However STAR will map to next exon(15bp), then softclip remaining bases, and no chimeric alignment output, lead to fusion false negative.

b8a432237696e9a8819eaca7f3754742

If I increase alignSJDBoverhangMin to 18bp, I will get a correct chimeric alignment. But alignSJDBoverhangMin=18 is too big, lead to a lot of chimeric alignment between adjacent exons.

I think this can be avoid by increase the penalty for a mismatch. But STAR do not have a option to set it. Can I have a way to adjust it(e.g. change the code if it is written as a const value)?

You can adjust the total number of mismatches allowed per read --outFilterMismatchNmax, but there is no way to limit mismatches per fragment.

I think this maybe hard to adjust, so I will close this issue.