GWAS QC step - Heterozygosity

Question

[jun0914]

Hi, I'm doing QC step with genetic data before doing imaging genetics study.

I use plink version 1.09.

I calculate heterozygosity rate to exclude individuals that has 3SD from mean value.

By calculating (N(NM)-O(HOM))/N(NM), I was able to get 'het', which is heterozygosity rate. The result is below table.

I filtered 3SD away from mean value, and this sorted out 113 subjects.

But I realized that each population(White,Black,Hispanic,Asian,Others) have different distribution of heterozygosity rate clustered, and about 80 people of excluded subjects were Asian.

Here are my questions.

1. Do I have to seperate population before performing any QC?
2. If not, do I have to just remove 80 asian, which is about half of full asian population?

Thank you.

Answer 1

[chrchang523]

1. When performing a QC step that uses allele frequencies, yes, it might be useful to split by population. However, that isn't a fully general solution: what do you do about people with mixed ancestry? So it's reasonable to just use a looser QC cutoff instead.
2. Your instincts are sound: it does not make sense to throw out half of your asians. If you aren't splitting by population, I'd go with a 6-10 SD cutoff; 3 SD is definitely too strict.

[jun0914]

Thanks a lot! I decided to go with less strict cutoff, and found out that I can save 100 subjects(including 80 asian) by applying 4SD cutoff instead. This sounds reasonable.