Add final filters on variant list

[ihodes]

For e.g., I want to get rid of all variants that have evidence of the variant allele appearing in the normal reads of, say, more than 1%. A lot of the variants I reject in manual review have a few reads showing the variant allele in normal.

Other conditions like overall errors in the reads, mapq, etc, might be work adding.

This would require some sort of integration with Vaxrank (or this could be a feature in Vaxrank, to keep things simpler; Vaxrank would then need to take Normal/Tumor BAMs as well, which might be a nice thing).

[iskandr]

Ha, Vaxrank will evolve into @timodonnell's Varlens read evidence filtering :-)

I think we should these things to Vaxrank if we want to keep the variants but down-weight those that fail certain rules. If we have rules for automatically dropping variants altogether then we should use Varlens or some similar tool.

[iskandr]

One pattern I've observed is reads with inconsistent patterns of adjacent mismatches by a variant.

If every variant supporting read also contains one other mismatch consistently, then I would assume we're capturing two co-occurring variants. If, on the other hand, every read has different mismatches then I would assume that we called the variant by chance or at least that there is something difficult about the alignment region.