How Large Would Inbreeding Coefficient Be To Be Anomalous?

Question

[nnlnn]

I checked inbreeding coefficient (F) on my samples (around 200) using plink

plink --file mydata --het

and found the distribution of F pretty symmetric:

and the short summary:

> summary(het$F)
    Min.  1st Qu.   Median     Mean  3rd Qu.     Max. 
-0.35990 -0.07199  0.01568  0.02755  0.13460  0.40790

The F values are quite big (either positive or negative), which made me worry about sample quality. Presumably they are all unrelated. Can anyone tell me whether this distribution of F indicates some problems and what they are? Or, is there a rule of thumb about what F value is considered normal for unrelated, cleaned samples? Thanks!

Answer 1

[zx8754]

For relatedness try IBS/IBD estimation:

• plink 1.9 https://www.cog-genomics.org/plink/1.9/ibd
• old website: http://zzz.bwh.harvard.edu/plink/ibdibs.shtml

PI_HAT:

• Identical twins, and duplicates, are 100%identical by descent (Pihat 1.0)
• First-degree relatives are 50% IBD (Pihat 0.5)
• Second-degree relatives are 25% IBD (Pihat 0.25)
• Third-degree relatives are 12.5% equal IBD (Pihat 0.125).

[Marco]

Hi, I am wondering how to test if two individuals are 1st degree relatives or 2nd degree relatives if the Pihat is between 0.5 and 0.25.. Is there any statistical testing method? Hoping to hear from you

[nnlnn]

Thanks, but I am afraid I was asking a different question. The inbreeding coefficient is per-individual, not for pairs of individuals as IBD is.

[jxchong]

Inbreeding coefficient (F_ped) is the same thing as kinship coefficient between the parents.

Answer 2

[jxchong]

IMO you've got an enormous problem: F of 0.25 would be the offspring of a father/daughter or mother/son or brother/sister pairing. Even in the data I've seen (for a founder population), F calculated by PLINK is <50% of your max and min values.

This distribution of F values indicates problems (both extreme excess of homozygosity and extreme deficiency of homozygosity). The deficiency of homozygosity could be due to sample contamination (enriching for heterozygous sites). I'd have to think a little harder about the cause of the extreme excess of homozygosity.

Have you already QCed the data for the basics? You say "cleaned" but I'm not sure what you mean. Have you done sex check, removed SNPs that deviate excessively from HWE, remove SNPs with low CR, remove individuals with low CR? If you have a lot of bad SNPs that are biased towards homozygous calls, you could potentially generate data with excess homozygosity.