Confused about the AF tag in VCF from samtools/bcftools mpileup

Question

[Makplus T]

Hi,

I have called the mutation in Human NGS data follow the samtools/bcftools mpileup call, and got the VCF files. Due to the false positive, I need to go further filter, just like filter the position with more than 20 reads support, Or the remove the variants frequency less than threshold value, (eg: AF < 4% ).

The VCF file from bcftools call do not contain the INFO/AF tags, so I add it with below command:

 bcftools +fill-tags in.bcf -Ob -o out.bcf -- -t AF

But after AF added, I only got two kinds of AF values for my variants records, both AF =1 for GT 1/1, and AF=0.5 for GT 0/1.
It seems AF in samtools in not the same concept in GATK tools? with GATK I can get the AF for the site = alt reads / reads depth, and in samtools, I only got the AF value = AC/AN, here AC is always reported as 1 and AN is 1 or 2 in my samples.

May be some mistakes need to be pointed out?

Thanks

[Kevin Blighe]

It's possible that they are different. Check the VCF header to see how AF is defined there in your GATK-produced VCF compared to the mpileup-produced VCF.

[Makplus T]

Thanks Kevin, I think I got the point. AF in GATK-produced VCF stands for "allele fraction of alternating allele in tumor", while AF in VCF produced by bcftools means "allele frequency", I think this is a population concept, "allele fraction of alt allele in all samples" . So when I only give one human sample, it always reports AF=1 or AF=0.5.

Answer 1

[Renesh]

The AF in bcftools and GATK should be the same. AF is calculated as AC/AN, where AC is total ALT allele count and AN is total allele called in genotypes. If you have only one genotype or sample, AF = 1 for GT = 1/1 refers to you have 2 ALT allele and the total is also 2 alleles. Similarly, AF=0.5 for GT 0/1 refers to 1 ALT allele and a total of 2 alleles. If possible, please share a few lines of data from your VCF file for further clarification.

[Makplus T]

Makes sense. I provide single sample to samtools mpileup , so I think AF = 0.5 or 1 is the actually the right result.