hello,

I want to extract the genotype of the germline variant. As it is a human sample, it is expected to be diploid (2n).

In the example below, three variants were observed at one SNP site, and an allele with a high number of reads among the three variants was selected, and sample#1 shows 0/2 and sample#2 shows a 1/2 genotype.

However, we want to use only alleles with AD > 100, and we want to proceed with genotype calling for alleles with AD > 100. As a result, the genotypes of sample#1 and #2 we want are both 2/2. (When AD<100 was simply applied, it was called 0/0. It seems that additional filtering is needed.)

#CHROM  POS     ID      REF     ALT     QUAL    FILTER  INFO    FORMAT  sample#1   sample#2
rs00001        174     .       A       T,G     46256.89        PASS    AC=1,2;AF=0.250,0.500;AN=4;BaseQRankSum=3.60;DP=22344;ExcessHet=3.0103;FS=0.000;MLEAC=1,2;MLEAF=0.250,0.500;MQ=60.00;MQRankSum=14.12;QD=14.52;ReadPosRankSum=4.62;SOR=1.350     GT:AD:DP:GQ:PL  0/2:7,4,2786:2808:22:38023,37997,38116,0,76,22  1/2:1,10,377:391:99:8253,8071,8052,195,0,482

Additionally, after applying AD> 100 filter, we want to consider the balance between heterozygous alleles (AB).

If  AD().1/DP > 0.7,  homozygous ref allele is called 
If  AD().1/DP < 0.3,  homozygous alt allele is called 
If  0.3 <AD().1/DP <0.7,  both ref and alt alleles are called (heterozygous)

In the example below, genotype of both sample #1 and #2 should be called as 1/1.

#CHROM  POS     ID      REF     ALT     QUAL    FILTER  INFO    FORMAT  sample#1   sample#2
rs00002       149     .       G       A       176718.50       PASS    AC=2;AF=0.500;AN=4;BaseQRankSum=1.76;DP=10932;ExcessHet=4.7712;FS=0.000;MLEAC=2;MLEAF=0.500;MQ=60.00;MQRankSum=3.000e-03;QD=16.18;ReadPosRankSum=4.69;SOR=0.693 GT:AD:DP:GQ:PL  0/1:1182,4788:5974:99:100100,0,11535    0/1:1263,3687:4950:99:76628,0,17056

I have considered several things, but I am asking a question because it is difficult to apply a filter. Besides variantfiltration, you can use other tools (selectvariant, vcffilter, bcftools, snpsift). Thank you for reading and ask for an answer if anyone knows the answer.