0
$\begingroup$

I'm seeing output from PBWT that looks like this:

3   112841  rs78923776  C   G   .   PASS    RefPanelAF=0.000153988;AN=2;AC=0;INFO=1 GT:ADS:DS:GP    0|0:0,0:0:1,0,0
3   112841  rs78923776  C   T   .   PASS    RefPanelAF=0.078534;AN=2;AC=1;INFO=1    GT:ADS:DS:GP    0|1:0,1:1:0,1,0

The first line says that my genome is CC at this position, but the second line (for the second alt allele) says that my genome is CT at this position. OK, the first line couldn't say this, so it calls me as REF/REF, but this call has to be interpreted in the context of the second call, and can't be taken at face value.

That's why I wonder if this is valid VCF or if it's actually breaking the specification.

The line should be 'squashed' down to the following (I think):

3   112841  rs78923776  C   G,T .   PASS    RefPanelAF=0.000153988,0.078534;AN=3;AC=0,1;INFO=1  GT:ADS:DS:GP    0|2:0,0:0:1,0,0
$\endgroup$
2
  • 2
    $\begingroup$ Many pieces of software can only work with biallelic diploid calls, so they have to split up multiallelic genotypes into separate lines. $\endgroup$
    – user438383
    Mar 9 at 13:02
  • $\begingroup$ Thanks. Is there a standard way to work out the genotype call? $\endgroup$
    – Dan Bolser
    Mar 10 at 14:45

1 Answer 1

2
$\begingroup$

It is a valid multi-allelic entry, and most variant callers produce such entries. However, comparison between two such gets more complicated. With biallelic entries, 2 CHR-POS-REF-ALT entries either match or they don't. With multi-allelic entries, they can match completely, partially or not at all, which means you also need to be able to specify how much of a match is a partial match etc. This is a hindrance in annotation processes that annotate using the CHR-POS-REF-ALT fields.

This also introduces cases where an indel can become a multi-allelic entry with a nearby SNV like so:

chr1    100    GA    G
chr1    101    A    AA

can now be represented as

chr1    100    GA    G,AA

which can get confusing. Operations are always more predictable when VCF has been decomposed and left-aligned.

$\endgroup$
6
  • $\begingroup$ Good to know this is valid. I'm having trouble interpreting some of the genotypes I'm seeing. Is there a standard way to do this? $\endgroup$
    – Dan Bolser
    Mar 10 at 14:44
  • $\begingroup$ Can you give me an example of the difficulty you're facing? $\endgroup$
    – Ram RS
    Mar 10 at 18:42
  • $\begingroup$ Hi @Ram, there are several genotype call combinations that I don't know how to 'decode'. I've written up a report here: github.com/richarddurbin/pbwt/issues/52#issuecomment-1064125959 The simplest way I can describe it is to say that, if the call is split across two lines (for a tri-allelic variant) the REF part of the call should be the same on both lines, no? $\endgroup$
    – Dan Bolser
    Mar 14 at 14:27
  • $\begingroup$ I glanced through your thread there and I think you're in the gray area between biology and bioinformatics. VCF entries assume reference sequence is reality, so yes, a multi-allelic site, when split, will have the same REF bases. Assuming that the mutated site is still diploid, a tri-allelic variant will have GT 1/2 where 1 is ALT-1 and 2-ALT-2 (so no REF base) $\endgroup$
    – Ram RS
    Mar 14 at 17:07
  • $\begingroup$ Right, I'd put that in VCF as something like ./1 and ./1 (assuming I had to split alternative alleles out into separate bi-allelic lines for some reason). $\endgroup$
    – Dan Bolser
    Mar 16 at 14:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.