Multiallelic SNPs are called as bi-allelic

Hello!

Thanks for releasing and supporting pepper - such a useful tool!

I’ve run variant detection according to this manual for Plasmodium falciparum reference genome and ONT reads.

I know that there are 3 haplotypes in the sample, but the majority of variants from PEPPER were bi-allelic. For example, a clearly triallelic site with A/C/G supported by 32/148/82 reads, was called as bi-allelic: Pf3D7_08_v3 736271 . G C 38.4 PASS . GT:GQ:DP:AD:VAF:PL 0/1:34:245:72,143:0.583673:38,0,35

Tri-allelic calls were few, just 0.3% of PASS calls, and they were mostly low coverage variants like

Pf3D7_09_v3	13563	.	A	T,G	21.4	PASS	.	GT:GQ:DP:AD:VAF:PL	1/2:2:2:0,0,0:0,0:16,4,1,4,0,1

Is there a flag to force pepper to call multiallelic variants?

Thanks! Sergey

[kishwarshafin]

@naumenko-sa ,

Thank you for pointing out the possible use-cases. I have created an internal issue with a mid-priority level. As this is a long-term goal, I’d close this issue. If we get to this, someone from our team will update this issue.

[naumenko-sa]

Thanks, please update this issue if you ever have a chance to push the multiallelic use case!

There are several potential applications of multiallelic variant calling:

• population calling in humans, for example in Gnomad Genomes database 13% of variants are reported to be in multi-allelic sites: https://gnomad.broadinstitute.org/blog/2017-02-the-genome-aggregation-database/
• problems like mine, when were are trying to call haplotypes in a mixture of samples (bacterial, viral etc)
• plant genomics.