Variant calling for MNP

Question

[nilus1432]

Hi All,

I would like to know is there any program for variant calling for multiple-nucleotide polymorphisms. For example, detecting cases like, Reference = 'ATA', Sample = 'GTC'.

Thanks a lot

Answer 1

[Len Trigg]

RTG Core calls haplotypes such as this. In reply to rbagnall, the benefit of calling as haplotypes rather than as individual SNPs is that the variant caller is able combine the evidence across both sites rather than calling them independently (e.g. read evidence containing all of ATA, ATC, GTA, GTC looks like het A->G and het A->C when looking at the sites separately, but is completely bogus from a haplotype perspective).

Whether the output is represented as multi-base calls or is decomposed into smaller variants is orthogonal and largely personal opinion. Either way, with complex calls you are likely to encounter cases where variants may have multiple alternative, equivalent representations. Even decomposition tools don't get you to a canonical form for all cases, so haplotype-based variant comparison tools such as rtg vcfeval are critical for accurate comparison of complex variants.

[rbagnall]

Thanks Len, that makes sense.

Answer 2

[User 59]

Freebayes calls MNPs and complex polymorphisms as well as SNPs and indels

[rbagnall]

I'm curious,

What is the benefit of calling a multiple-nucleotide polymorphism as shown in @nilus1432 question, rather than calling an A>G and an A>C single nucleotide variant? The latter are easier to annotate (missense, frequency in population controls etc...).

[User 59]

Because you can annotate them correctly. Calling two adjacent SNPs and passing them to something like VEP will give you two different SNP effects, but if it's a genuine dinucleotide change within say a codon, you need to be able to assess the effect of their presence as a unit.

[rbagnall]

Thanks Daniel,

Yes that's true and something I hadn't considered. I will have to see if VEP annotates such MNPs; it looks like SNPEff does. Also, the train of thought led me to this paper which describes the issue a bit more for anyone interested, though I haven't tried the described software: http://www.biomedcentral.com/1471-2164/16/569