I've run into the following bug while running GenotypeGVCFs:
##### ERROR MESSAGE: cannot merge genotypes from samples without PLs; sample <ID redacted> does not have likelihoods at position 1:1115551
The input file in question is a gVCF produced by merging a large number of smaller gVCFs using CombineGVCFs (all tasks were run using version 3.1). What's happening is that the position 1115551 doesn't exist in that particular sample:
#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT <Sample_ID> 1 1115550 . AC A,<NON_REF> 118.73 . BaseQRankSum=-0.377;DP=15;MLEAC=1,0;MLEAF=0.500,0.00;MQ=60.72;MQ0=0;MQRankSum=-1.093;ReadPosRankSum=-0.811 GT:AD:DP:GQ:PL:SB 0/1:7,6,0:13:99:156,0,188,177,207,384:4,3,3,3 1 1115552 . C <NON_REF> . . END=1115552 GT:DP:GQ:MIN_DP:PL 0/0:15:0:15:0,0,31
But when the sample is combined with other samples, that position gets filled in with a simple "0/0", without any PLs (or any of the other fields, including AD, DP, GQ, etc.), which causes the GenotypeGVCFs to choke.
I can imagine there might be other scenarios that will result in a "0/0" genotype field, so perhaps the easiest way to fix this would be to make sure that any "0/0" actually gets output as "./.:.:.:.:.".