-
Notifications
You must be signed in to change notification settings - Fork 260
Closed
Description
I was collecting some mendelian error rate stats on a multi-sample VCF containing fairly diverged trios. One "issue" I noticed was that many trios would have homozygous-reference genotypes for all three members of that trio. Since the samples could be fairly diverged, sometimes up to half of the "ngood" variants are these technically mendelian-consistent but otherwise incredibly boring records.
I tried adding a flag and if statement to skip any variants where there was no variation within that trio, and that seemed to work, placing this
if ( args->filter_reference && mom==1 && dad==1 && kid1==1 && kid2==1) continue;
right before
Line 651 in f313599
| if ( (kid1&dad && kid2&mom) || (kid1&mom && kid2&dad) ) { ret |= HAS_GOOD; trio->stats.ngood++; continue; } // both children's alleles phased |
The current behaviour is perhaps the most valid, but not necessarily the most useful for some purposes, so curious for additional thoughts.
Best,
Alex
Metadata
Metadata
Assignees
Labels
No labels