hemizygous variants didn't get picked

[lindakjcao]

Hi,

We have two hemizygous variants failed to be picked by the silver cmd line below:

  Silver '--group-expr "proband:'+
      'proband_has_variant(kid) && '+
      'variant.FILTER == \'PASS\' '+
      '&& \'FT\' in kid && kid.FT == \'PASS\' '+
      '&& variant.QUAL >= 30"'

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:.

[brentp]

proband_has_variant is not in the js that I distribute with slivar, so you'll have to share that code.

[lindakjcao]

Thank you so much for the quick response. Sorry about that. Here is the code:
function proband_has_variant(proband) {
return proband.alts > 0
}

[brentp]

slivar is most tested on diploid variants but I think this should work.
Are you using the latest version?

[lindakjcao]

we used v0.2.7. Probably not the latest one.

[brentp]

please try with latest then we will check from there.

[lindakjcao]

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?
chrX 154535277 rs1050829 T C 80.10 PASS AC=1;AF=1.000;AN=1;DP=10;FS=0.000;MQ=250.00;QD=8.01;SOR=0.693;DB GT:AD:AF:DP:GQ:FT:F1R2:F2R1:PL:GP 1:0,10:1.000:10:80:PASS:0,6:0,4:115,0:8.0096e+01,0.0000e+00

[brentp]

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.

[weixuanfu]

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.
proband.zip
trio.zip

1. I tested command below with silvar v0.3.0, 0 variant can pass

/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

3. However, still with inputs in proband.zip, I changed the expression as shown below, then 0 variant 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) && variant.FILTER == 'PASS' && 'FT' in kid && kid.FT == 'PASS' && variant.QUAL >= 30" \
  --out-vcf test_out.vcf

[weixuanfu]

Correction for test 3 above. It works with export SLIVAR_FORMAT_STRINGS=1

[weixuanfu]

I also tested duo family structure:
duo.nodad.zip
duo.nomom.zip

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

export SLIVAR_FORMAT_STRINGS=1
/mnt/isilon/dgd_public/clin-air/v2.0.0/tools/slivar/0.3.0/slivar expr \
  --pass-only \
  --vcf test.duo.nodad.vcf \
  --ped ped.duo.nodad.txt \
  --alias alias.duo.nodad.txt \
  --js filters.js \
   \
  --group-expr "proband:proband_has_variant(kid) && variant.FILTER == 'PASS' && 'FT' in kid && kid.FT == 'PASS' && variant.QUAL >= 30" \
  --out-vcf test_out.vcf

However
Test 5. duo for kid + dad (kid (GT=1) + dad (GT=0) in duo.nomom.zip) works with commands below, both variants can be picked up.

export SLIVAR_FORMAT_STRINGS=1
/mnt/isilon/dgd_public/clin-air/v2.0.0/tools/slivar/0.3.0/slivar expr \
  --pass-only \
  --vcf test.duo.nomom.vcf \
  --ped ped.duo.nomom.txt \
  --alias alias.duo.nomom.txt \
  --js filters.js \
   \
  --group-expr "proband:proband_has_variant(kid) && variant.FILTER == 'PASS' && 'FT' in kid && kid.FT == 'PASS' && variant.QUAL >= 30" \
  --out-vcf test_out.vcf

I guess that silvar may not handle haploid and diploid variants at the same time.

[weixuanfu]

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.

[weixuanfu]

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 proband_has_variant from

function proband_has_variant(proband) {
    return proband.alts > 0
}

to

function proband_has_variant(proband) {
    return proband.AD[1] > 0
}

the 2 variants can be picked in all the tests above

[brentp]

ok. I think that's a better way to go. alts really makes most sense for diploid variants.
also, I think you shouldn't use both --ped and --alias

[weixuanfu]

will silvar have better support for hemizygous genotypes? like alts=4 for hem-ref and alts=5 for hem-alt.

[brentp]

Yes, I committed last month code to handle accessing the genotype:
9f7d995

so you'll be able to do sample.GT which would be [0, 1] for a heterozygous allele, and [1] for a hemizyous variant.
I'll try to get this out next week.

[weixuanfu]

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:

(proband.alts > 0 || (proband.alts < 0 &&  proband.GT[0] == 1)

[brentp]

You can try it the attached
slivar_dev.gz

just download, gunzip, chmod +x and run as ./slivar_dev