AF = 0.5 in chrX, VCF of a male

Question

[paolo]

I am looking at a file VCF from whole exome sequencing of a male.

On chrX I find several variants with AC=1, AF=0.5, and AN=2.

How should I consider these variants?

[Dave Carlson]

It might be useful to plot some quality metrics for these SNPs in comparison with other non-heterozygous SNPs to see if they are aberrant in some way.

Do they by any chance fall within the pseudoautosomal regions of the chromosome?

[Pierre Lindenbaum]

How was the VCF called? With GATK? By default, GATK has no notion of sexual chromosome

[paolo]

I have not generated the VCF myself, but yes, it has been called with GATK 3.7. What does it mean? Why do I find some variants on ChrX that have AN=2 and AC=1 (an example below), while the remaining variants on ChrX have AN=2 and AC=2 (which would make more sense)?

chrX 13618410 . T C 52.78 PASS AC=1;AF=0.5;AN=2

[Pierre Lindenbaum]

a BAM is not perfect. If there is any region with a lot of mismatches (bad reads, duplicated regions, low complexity region...) , GATK will do is best to call the variant as diploid . See the FORMAT/AD field to get more information about the calls.

Again GATK treats all chromosomes as autosomal+diploid.

[paolo]

I have 248 calls on Chr X with AF = 0.5 (out of 1595 total calls on X). If we look at the FORMAT/AD field of ChrX with AF = 0.5, we see that both the alleles have been called. I paste here three calls, as an example, including the format field.

chrX 118605047.00 . G A 597.77 PASS AC=1 AF=0.5 AN=2 GT:AD:DP:GQ:PGT:PID:PL 0/1:66,17:83:99:0|1:118605001_G_T:626,0,3932

chrX 118605051 . A G 486.77 PASS AC=1 AF=0.5 AN=2 GT:AD:DP:GQ:PGT:PID:PL 0/1:65,15:80:99:0|1:118605001_G_T:515,0,3735

chrX 118605056 . C T 375.77 PASS AC=1 AF=0.5 AN=2 MLEAC=1 GT:AD:DP:GQ:PGT:PID:PL 0/1:60,12:72:99:0|1:118605001_G_T:404,0,3534

[Pierre Lindenbaum]

66,17 65,15 60,12

we expect 50%,50% for a nice Het... for 6 REF you have only one ALT.

[paolo]

So, can I consider calls like these of good quality and assume the REF allele?

[d-cameron]

Are these variants in the pseudoautosomal region?

[paolo]

My calls for chrX are all in the region between PAR1 and PAR2, so they are not in the pseudoautosomal region.

Answer 1

[abbey]

Do you have copy number information for this sample? In my cancer samples, it is common to have X duplications in XY-originating tumors that can cause something like this, which is why I do additional downstream analysis to consider copy number context in my variant calling.

I have a conflict of interest because this is my lab's software, but TitanCNA is a tool I use on a weekly basis for identifying variants using SNPs (from GATK or other variant callers) and copy number context (from ichorCNA) together: https://github.com/gavinha/TitanCNA

Answer 2

[Jeremy]

AN is the number of alleles at that position, and AF is the allele frequency, so I would interpret your data as meaning that your subject is heterozygous at that position. See the VCF Specification linked below:

VCF

[paolo]

But it is in chrX, and the subject is a male.

[karl.stamm]

according to the software, het just means one copy. it doesn't know that one copy is normal for this individual.

[paolo]

I see, but most of the variants on ChrX are AC=2; AF=1; AN=2