Closed sooheelee closed 6 years ago
In addition, we should be able to supply the GRCh38.fasta.alt file to our tools so GATK can intelligently handle the ALT contigs. That is, the program will know that reads mapped to chr1 or chr1_xyz_ALT are in fact on the same molecular chromosome (contig).
Making the callers ALT aware is a valid feature request -- but it's a big one. You'll want to open a separate issue ticket for that.
Ok.
@sooheelee Done in PR #4466 after verifying that it doesn't hurt sensitivity on hg19.
@vdauwera we should modify the M2 WDLs. @davidbenjamin this will improve your sensitivity.
Currently Mutect2 uses the MateOnSameContigOrNoMappedMateReadFilter filter that filters out any paired read whose mate maps to a different contig. This filter, if I recall correctly, used to be the hidden filter in HaplotypeCaller code that could not be turned off. It necessitated that I remove 0x1 flags in the GRCh38 tutorial (see section 6.1 of https://gatkforums.broadinstitute.org/gatk/discussion/8017/) so as to be able to call variants associated with a sample with an alternative haplotype.
This filter is now exposed so that users can disable it. In addition to disabling this filter for ALT-aware data, I recommend we turn it off by default for somatic analyses, for any reference. This allows us (i) to call on ALT-aware mappings if data is such and (ii) call on SNPs and indels generated by putative structural variants that go across contigs.
I know that this filter is active in the GATK4.beta.3-Mutect2 (see last line):
Also, can we make the standout that counts filtered reads more human readable please.