Heterozygosity by individual

Question

[Li]

How to compute heterozygosity for each individual?

Currently I use:

plink --bfile file --hardy --out file --noweb

but I need .hwe files for each individual separately in a dataset of hundreds of individuals.

Answer 1

[chrchang523]

--hardy requires many individuals to judge deviation from Hardy-Weinberg equilibrium. Single-individual .hwe files will not be useful.

If you are trying to perform quality control on samples, consider plotting top principal components (computed with EIGENSOFT 6, or plink --pca) and removing extreme outliers.

[Li]

I need to compute heterozygosity per individual from genotype data. Not for quality control.

[chrchang523]

Ah. --het (https://www.cog-genomics.org/plink2/basic_stats#ibc ) should come in handy, then.

[Li]

Thanks. Two problems:

1. --het computes observed and expected autosomal homozygous genotype counts for each sample. However, I need heterozygosity for the X chromosome too (separately). --chr 23 --het gives zero o(hom) and e(hom). My sample is females only.

2. --het gives an observed heterozygosity of 40% for my example individual, while if I calculate the average o(het) of autosomal snips of the same individual from a .hwe file, o(het) is 30%. For expected heterozygosity the numbers are 36% vs 16%. Thus, the two methods yield very different answers, although if I understand correctly, they should measure the same thing.

As a context: the goal is to compare heterozygosity of X and autosomes between groups, and in order to do a t-test, I need X and autosomal het per individual. I already have average X and autosomal het per group.

[chrchang523]

It's a dirty hack, but if your .bim file has numeric chromosome codes, you can force the X chromosome to be treated like an autosome by specifying a species with more chromosomes (e.g. "--dog").

[Li]

That works, but I still need to compute this using --hardy too, per individual. Creating a text file with one person, and using --keep myfile.txt is not possible because of the high number of individuals.

Answer 2

[Johan Zicola]

VCFtools has the function --het such as vcftools --vcf input.vcf --het --out output.het

--het

Calculates a measure of heterozygosity on a per-individual basis. Specfically, the inbreeding coefficient, F, is estimated for each individual using a method of moments. The resulting file has the suffix ".het".

Check the full manual: https://vcftools.github.io/man_latest.html

Check this post if you have difficulty to interpret the results: Is the heterozygosity flag (--het) in vcftools calculate observed and expected heterozygosity?