If you happen to see a question you know the answer to, please do chime in and help your fellow community members. We appreciate your help!
Test-drive the GATK tools and Best Practices pipelines on Terra
Check out this blog post to learn how you can get started with GATK and try out the pipelines in preconfigured workspaces (with a user-friendly interface!) without having to install anything.
0 depth but very high genotype call quality
This variant really puzzles me:
chr20 29651901 . T C 64346.73 PASS AC=35;AF=0.500;AN=70;BaseQRankSum=3.821;ClippingRankSum=-0.329;DP=131;FS=2.944;InbreedingCoeff=-1.0000;MLEAC=35;MLEAF=0.500;MQ=52.88;MQ0=0;MQRankSum=-2.724;QD=27.79;ReadPosRankSum=3.747;VQSLOD=0.368;culprit=ReadPosRankSum GT:AD:DP:GQ:PL 0/1:5,1:6:99:2348,0,1280 0/1:1,2:3:99:1864,0,350 0/1:.:.:33:2587,0,33 0/1:0,0:0:99:903,0,191 0/1:0,1:1:99:1530,0,601 0/1:3,0:3:99:2725,0,686 0/1:2,0:2:99:2642,0,504 0/1:.:.:99:2158,0,250 0/1:1,0:1:99:1344,0,726 0/1:4,0:4:99:1961,0,808 0/1:2,3:5:99:1275,0,758 0/1:6,1:7:99:2394,0,854 0/1:5,1:6:99:1610,0,374 0/1:6,1:7:99:1772,0,1712 0/1:1,1:2:99:1682,0,391 0/1:1,2:3:99:1693,0,625 0/1:5,2:7:99:1165,0,739 0/1:1,3:4:99:2155,0,389 0/1:1,3:4:99:2034,0,517 0/1:2,0:2:99:1638,0,345 0/1:1,0:1:99:1712,0,292 0/1:6,0:6:99:1924,0,691 0/1:1,3:4:99:2186,0,472 0/1:1,0:1:99:1621,0,543 0/1:2,0:2:99:853,0,513 0/1:1,1:2:99:2010,0,632 0/1:7,4:11:99:2132,0,1699 0/1:1,0:1:99:1856,0,303 0/1:3,1:4:99:1359,0,1002 0/1:0,1:1:99:990,0,655 0/1:3,1:4:99:2668,0,436 0/1:4,2:6:99:2466,0,601 0/1:3,1:4:99:2301,0,427 0/1:2,0:2:99:1384,0,546 0/1:5,0:5:99:1523,0,751
The call was produced with GATK 2.7.2 HaplotypeCaller from 35 exomes and recalibrated with VQSR. As you can see the loci is covered pretty badly. Despite this, the genotype call quality are all 99 for all samples. Even worse, '0/1:0,0:0:99' should mean that 0 coverage thus still heterozygous call with very high confidence on the call. How could this be? And when I visualize the actual alignment of the loci for the 0 covered, I did see 8 reads mapped to the position with mapping quality ranging from 25 to 60 (BWA mapped) and no variant allele was seen. Please help, what could trick the VQSR process and leave such calls so believable? Should we still apply a coverage filter after VQSR, I think this is not ideal and exactly what we try to avoid by using VQSR!