RefCall vs. PASS - Githubissues

kishwarshafin / pepper

PEPPER-Margin-DeepVariant

MIT License

245 stars 42 forks source link

RefCall vs. PASS #160

Closed yifeng-evo closed 2 years ago

yifeng-evo commented 2 years ago

Hi,

I am not sure why some positions are considered as "refcall", which indicates the "call is homozygous". When I checked the position in IGV, it seems that it is a heterozygous position. I was wondering how refcall is different from PASS. For example： NODE_10237_length_5294_cov_2.741049 216 . A AT 17 PASS . GT:GQ:DP:AD:VAF:C 0/1:17:36:12:0.333:P NODE_10237_length_5294_cov_2.741049 239 . T A 0 refCall . GT:GQ:DP:AD:VAF:C 0/0:21:36:11:0.305556:DV I am new to VCF files so maybe I didn't understand the format correctly. Could you please help me to clarify this? Thank you!

kishwarshafin commented 2 years ago

@yifeng-evo ,

The variant in position 216 is a PASS variant with GT of 0/1 which means it's a het variant.
The variant in position 239 is a refCall meaning it's likely an error in sequencing platform and assigned GT is 0/0 meaning it's homozygous to the reference.

yifeng-evo commented 2 years ago

It seems to be not a homozygous position in IGV (A in green and T in red). I checked a couple of other refcall positions and they also look similarly. Is it because some reads were filtered due to bad quality? or maybe I used "--pepper_min_coverage_threshold 20" so one allele could be considered as homozygous if alt is supported by <20 reads?

kishwarshafin commented 2 years ago

@yifeng-evo ,

There are a variety of things that go into the determination of a site being a true variant or not. Including base_quality, mapping_quality, strand_distribution etc. What it means here is the DeepVariant looked at this site and decided it is not a variant.

The --pepper_min_coverage_threshold 20 parameter only affects candidate finding. As you are seeing the variant in the VCF, it means that didn't affect your calling.