iGenVar - Find SNPs & Indels modeled on GATK

[Irallia]

Call SNPs & Indels: Since GATK is currently the best standard, we want to use these methods.
GATK calls SNPs & Indels via Mutect2 and HaplotypeCaller.
They both have in common a local assembly and realignment of active sites.
Mutect2 is a somatic caller and, on the other hand, HaplotypeCaller is a germline caller.
Mutect2 uses the GATK tool FilterMutectCalls after the realingment, to filter somatic variants (opposed to germline variants), sequencing errors, ...
It looks like it is the best way to translate the tool HaplotypeCaller, written in Java, into C ++ for this purpose.

-> generating candidate haplotypes
-> local realignment using the pair HMM Model against the candidate haplotypes -> matrix of likelihoods for each read
-> local assembly: assemble these window aligned reads into an assembly graph of local variation
-> infer variants from assembled haplotypes: "Despite its name, HaplotypeCaller does not actually call haplotypes. Rather, it generates haplotypes as an intermediate step to discover variants at individual loci. Here we describe how the GATK engine determines which alt alleles exist in locally assembled haplotypes." (-> variant qualtiy score model)

Other links:

• Mutect2
• Talk about GATK HaplotypeCaller methods (PDF Slides)

Issue sketches:

• reimplement the local assembly and realignment in C++. This is described in: https://www.biorxiv.org/content/10.1101/861054v1.supplementary-material
• reimplement FilterMutectCalls in C++
• ...

Paper and Articles:

• Calling Somatic SNVs and Indels with Mutect2
• Somatic calling is NOT simply a difference between two callsets
• Pair HMM probabilistic realignment in HaplotypeCaller and Mutect PDF
• Description of alignment method and GPU implementation Link