-
Notifications
You must be signed in to change notification settings - Fork 23
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
hemizygous variants didn't get picked #154
Comments
|
Thank you so much for the quick response. Sorry about that. Here is the code: |
slivar is most tested on diploid variants but I think this should work. |
we used v0.2.7. Probably not the latest one. |
please try with latest then we will check from there. |
Hi, we found another case that has same hemi variant at 154535277, it was picked by Silvar. The only difference is the case is a singleton. So hemizygosity seems not to be the cause but something else. Any suggestion? |
can you try with latest version? if the results are not what you expect, then send a valid VCF with the 2 variants and the VCF header and I will take a look. |
Hi @brentp thank you for replying this issue so quickly. I made some mocked files to reproduce this issue. The 1st set is a trio (attached trio.zip) that mentioned above.
/tools/slivar/0.3.0/slivar expr \
--pass-only \
--vcf test.vcf \
--ped ped.txt \
--alias alias.txt \
--js filters.js \
\
--group-expr "proband:proband_has_variant(kid)" \
--out-vcf test_out.vcf 2 I tested command below with inputs in proband.zip (which only has proband's genotype), 2 variants can pass /tools/slivar/0.3.0/slivar expr \
--pass-only \
--vcf test.single.vcf \
--ped ped.single.txt \
--alias alias.single.txt \
--js filters.js \
\
--group-expr "proband:proband_has_variant(kid)" \
--out-vcf test_out.vcf
|
Correction for test 3 above. It works with |
I also tested duo family structure: Test 4. duo of kid + mom (kid (GT=1)+mom (GT=0/1) in duo.nodad.zip) with commend below, silver cannot pick the 2 variants
However
I guess that silvar may not handle haploid and diploid variants at the same time. |
I have made another duo with kid + mom but only change kid's GT from 1 to 1/1 in vcf (kid's GT = 1/1 and mom' GT = 0/1), then slivar can pick both variants when there is no mix of haploid and diploid GT in vcf. |
I checked the source codes about genotypes. the kid.alts should be -1 in this mix type GT in vcf. So if I change the function
to
the 2 variants can be picked in all the tests above |
ok. I think that's a better way to go. alts really makes most sense for diploid variants. |
will silvar have better support for hemizygous genotypes? like alts=4 for hem-ref and alts=5 for hem-alt. |
Yes, I committed last month code to handle accessing the genotype: so you'll be able to do |
Thank you! Could you please let me know how to compile the beta version of silvar from source codes? I need it to test the logic below:
|
You can try it the attached just download, gunzip, chmod +x and run as ./slivar_dev |
Hi,
We have two hemizygous variants failed to be picked by the silver cmd line below:
It seems the logic probing_has_variant(kid) doesn't handle this situation when GT = 1 (hemizygous). Could you please take a look? The version of Silver is v0.2.7. Thank you. Here are the variants:
#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT kid mom dad
chrX 154535277 rs1050829 T C 199.46 PASS AC=2;AF=0.500;AN=4;DP=68;FS=0.000;MQ=250.00;MQRankSum=4.951;QD=3.50;ReadPosRankSum=1.052;SOR=0.521;DB GT:AD:AF:DP:GQ:FT:F1R2:F2R1:PL:GP:PP:DN 1:0,23:1.000:23:119:PASS:0,13:0,10:154,0:1.1923e+02,0.0000e+00:154,0:Inherited 0:21,0:0.000:11:99:PASS:.:.:0,384:.:0,258:. 0/1:18,16:0.471:34:48:PASS:7,6:11,10:85,0,50:4.9792e+01,6.7304e-05,5.3000e+01:130,0,53:.
chrX 154536002 rs1050828 C T 186.46 PASS AC=2;AF=0.500;AN=4;DP=67;FS=7.399;MQ=250.00;MQRankSum=4.540;QD=3.39;ReadPosRankSum=3.433;SOR=1.302;DB GT:AD:AF:DP:GQ:FT:F1R2:F2R1:PL:GP:PP:DN 1:0,19:1.000:19:107:PASS:0,7:0,12:141,0:1.0672e+02,0.0000e+00:141,0:Inherited 0:21,0:0.000:11:99:PASS:.:.:0,384:.:0,258:. 0/1:18,18:0.500:36:48:PASS:8,8:10,10:85,0,50:5.0000e+01,6.5233e-05,5.3000e+01:130,0,53:.
The text was updated successfully, but these errors were encountered: