Incorrect number of FORMAT/PL values in .g.vcf

[dennishendriksen]

I'm running into an issue where Clair3 produces invalid .g.vcf output resulting in errors when used with downstream tooling, e.g. BCFtools:

Incorrect number of FORMAT/PL values at chr2:14862711, cannot merge. The tag is defined as Number=G, but found
  4 values and 3 alleles. See also http://samtools.github.io/bcftools/howtos/FAQ.html#incorrect-nfields

Example output (unfortunately I can't share the input due to its sensitive nature):

pileup.vcf.gz:

chr2    14862711        .       C       .       7.03    RefCall P       GT:GQ:DP:AD:AF:PL       0/0:7:2:1:0.5000:2970

full_alignment.vcf.gz:

chr2    14862711        .       C       N       7.97    PASS    F       GT:GQ:DP:AD:AF:PL       0/1:7:3:1,1:0.3333:2970

merge_output.vcf.gz:

chr2    14862711        .       C       N       7.97    PASS    F       GT:GQ:DP:AD:AF:PL       0/1:7:3:1,1:0.3333:2970

output.g.vcf.gz:

chr2    14862711        .       C       N,<NON_REF>     7.97    PASS    F       GT:GQ:DP:AD:AF:PL       0/1:7:3:1,1,0:0.3333:2970,990,990,990

I think the error is in the PL computation in case the alternate allele value is N, caused by https://github.com/HKU-BAL/Clair3/blob/v1.0.0/clair3/CallVariants.py#L1395? Since the VCF field PL is of type G the field still must have one value for each possible genotype according to the VCF specification.

Possibly the following was intended in full_alignment.vcf.gz?

chr2    14862711        .       C       N       7.97    PASS    F       GT:GQ:DP:AD:AF:PL       0/1:7:3:1,1:0.3333:990,990,990

[aquaskyline]

To be fixed in the next release.

[zhengzhenxian]

@dennishendriksen Fixed in v1.0.1, pls kindly try, thanks!