This site is now read-only. You can find our new documentation site and support forum for posting questions here.
Be sure to read our welcome blog!
Uncorrect Strand Bias due to Downsampling (HaplotypeCaller)
Hi GATK team,
Again thanks a lot for the wonderful tools you're offering to the community.
I have recently switched from UnifiedGenotyper to Haplotype Caller (1 sample at a time, DNASeq). I was planning to use the same hard filtering procedure that I was using previously, including the filter of the variants with FS > 60.
However I am facing an issue probably due to the downsampling done by HC.
I should have 5000 reads, but DP is around 500/600 which I understood is due to downsampling (even with -dt NONE). I did understand that it does not impact in the calling itself. However it is annoying me for 2 reasons
1) Calculating frequency of the variant using the AD field is not correct (not based on all reads)
2) I get variants with FS >60 whereas when you look at the entire set of reads, there is absolutely no strand bias.
Example with this variant
chr17 41245466 rs1799949 G A 7441.77 STRAND_BIAS; AC=1;AF=0.500;AN=2;BaseQRankSum=7.576;DB;DP=1042;FS=63.090;MLEAC=1;MLEAF=0.500;MQ=60.00;MQRankSum=0.666;QD=7.14;ReadPosRankSum=-11.896;SOR=5.810 GT:AD:GQ:PL:SB 0/1:575,258:99:7470,0,21182:424,151,254,4
When I observe all reads I have the following counts, well shared on the + and - strands
Allele G : 1389 (874+, 515-)
Allele A : 1445 (886+, 559-)
Could you please tell me how to avoid such an issue ? (By the way, this variant is a true one and should not be filtered out).
Thanks a lot.