Hi everyone,
I'm calling SNPs with UnifiedGenotyper:
java -Xmx2g -jar /Volumes/DroboLeandro/Analysis/Software/GenomeAnalysisTK/GenomeAnalysisTK.jar -T UnifiedGenotyper -R /Volumes/DroboLeandro/Analysis/Genomes/Reference/NCBIM37.chr.fa -I BALBcByJ.recal.bam --dbsnp /Volumes/DroboLeandro/Analysis/dbSNP/dbSNP.vcf --genotype_likelihoods_model SNP --metrics_file BALBcByJ.metrics.snp.txt -stand_call_conf 30 -stand_emit_conf 30 -o BALBcByJ.snp.raw.vcf -log BALBcByJ.UnifiedGen.snp.log
It appears to work fine, without any apparent error. But when I try to run VariantAnnotator on the raw.vcf file output, it gives an error because the VCF file is malformed:
ERROR MESSAGE: The provided VCF file is malformed at approximately line number 237: there aren't enough columns for line 0;FS=0.000;HRun=0;HaplotypeScore=0.0000;MQ=60.00;MQ0=0;QD=33.90 GT:AD:DP:GQ:PL 1/1:0,11:11:33.10:406,33,0
When I look at the region on the VCF, this is what I actually have:
chr4 58379627 rs32247182 A G 1694.62 . AC=2;AF=1.00;AN=2;DB;DP=57;Dels=0.00;FS=0.000;HRun=0;HaplotypeScore=0.8941;MQ=58.03;MQ0=0;QD=29.73 GT:AD:DP:GQ:PL 1/1:0,56:57:99:1728,162,0
chr4 58379655 rs32566361 T C 3398.59 . AC=2;AF=1.00;AN=2;DB;DP=100;Dels=0.00;FS=0.000;HRun=0;HaplotypeScore=1.8662;MQ=58.65;MQ0=0;QD=33.99 GT:AD:DP:GQ:PL 1/1:0,98:100:99:3399,277,0
0;FS=0.000;HRun=0;HaplotypeScore=0.0000;MQ=60.00;MQ0=0;QD=33.90 GT:AD:DP:GQ:PL 1/1:0,11:11:33.10:406,33,0
chr4 58380476 rs32067653 G T 1272.97 . AC=2;AF=1.00;AN=2;DB;DP=39;Dels=0.00;FS=0.000;HRun=0;HaplotypeScore=0.9987;MQ=58.92;MQ0=0;QD=32.64 GT:AD:DP:GQ:PL 1/1:0,39:39:99:1306,114,0
So UnifiedGenotyper is outputting incorrect VCF files with some incomplete lines (the one in bold), where only annotations are present. It is the 5th line, today, in completely independent samples. If I remove these incomplete lines, the downsteam tools will work just fine! But I'm affraid I'm loosing SNPs. Is it a bug on UnifiedGentoyper? Has anyone had the same error?
Thanks, Leandro
Hmm... maybe nothing is lost. Are all of the HaplotypeScores = 0.00000. It sounds like a bug, but maybe it only effects calls with 0 confidence?