Closed gcv closed 5 years ago
For @aneilbaboo: Change the validation so that end
is optional if readFail
(a new data point) is true (it should default to false).
For @gcv: Change the bioinformatics validation to look for filter
set to .
, and when that happens, set readFail
to true
.
I removed end
from VariantCall.
I looked more closely at VCF and determined that "end" is not required and the way we were using it is ambiguous. As it turns out, the userId
plus chromosome:version:start:sampleType:sampleId
is sufficient to uniquely identify a variantCall. I'm dropping "end".
The reason for this is that VCF represents actual "calls" of variants, not raw potentially overlapping sequence data. So the software that produces the VCF has resolved overlaps and presents a single call at a position.
@gcv - my changes require a small change in the bioinformatics code base.
The genotypeLikelihood
param is now pluralized: genotypeLikelihoods
. It is an array of numbers of a defined length. The length, it turns out, is dependent on the number of altBases
. I wrote up some documentation about in a code comment in a PR I am about to push. We can discuss.
Apart from changing genotypeLikelihood
=> genotypeLikelihoods
, you shouldn't have to do anything else.
Excuse me - there are three changes you'll need to make, the others being injecting a couple of booleans: readFail
and imputed
.
I'll put these in a separate issue.
Failing input:
Error:
This doesn't make sense because
variant-call/models.ts
does not sayend
is required, and noend
argument is passed in. It just saysend: Joi.number().min(Joi.ref('start'))
, which suggests it just needs to be at least as large asstart
.This variant is then imputed, which seems to validate just fine: