Hello everyone,
I've been Googling for a couple days now without success. The situation is like so:
114 human WGS samples (probands, i.e. they should be wholly unrelated). Three of them are significantly contaminated according to VerifyBamID2, with FREEMIX values of 0.24, 0.3, and 0.4. These three samples are estimated by KING (through PLINK2 --make-king-table) to be related at the 1st or 2nd degree level to ALL other samples. The 114 samples are made up of at least 4 independent sequencing batches.
I checked heterozygosity levels with PLINK2 --het, and the three mentioned samples have significantly increased heterozygosity. The most contaminated sample with FREEMIX=0.4 has almost half the number of HOM sites compared to the average sample in my N=114 cohort.
I'm trying to find evidence that heterozygosity inflates the KING estimate between unrelated individuals, but failing.
Does anyone here know anything about this? I want to understand what has happened.
Thanks in advance!
Yes, inflated heterozygosity inflates KING kinship coefficients. See equation 9 from the KING paper (https://pmc.ncbi.nlm.nih.gov/articles/PMC3025716/ ), where the kinship coefficient numerator is written as the number of both-heterozygous genotypes minus twice the number of opposite-homozygous genotypes.
Oh, indeed. They even write it out a few sentences later: "When the violation of HWE is in the direction of too little homozygosity (i.e. excessive heterozygosity), the robust estimator (Equation 9) can over-estimate the kinship coefficient."
Thank you very much, chrchang. That's one headache fewer. Yay.
The answer is yes, thanks to chrchang, the original KING paper explicitly says:
"When the violation of HWE is in the direction of too little homozygosity (i.e. excessive heterozygosity), the robust estimator (Equation 9) can over-estimate the kinship coefficient."
Log in to answer this question.
"The most contaminated sample with FREEMIX=0.4 has almost half the number of HOM sites compared to the average sample"
sorry, you mean half the number of HOM_REF or HOM_ALT sites?
Hello Jeremy, the output of PLINK2 --het does not specify this. I suppose it's the sum of both.