I found that several papers describe SNPs as a binomial distribution with the probability of "success" equals to minor allele frequency.

However, in my experiments, when I generate SNP array following this distribution, the simulation results behave very differently from the SNP array generated following the random mating procedures. I feel like there is some missing information about the description of a binomial distribution.

I wonder what will be a better mathematical description of the SNP array data? Besides a direct answer, any thoughts, or any suggestion of papers are also highly welcome.


Thanks for these suggestions, after reading relevant materials, I think what I observe is about the Hardy-Weinberg equilibrium affected by evolutionary influences like mutation and migration. I will continue to look and list findings here if I find anything interesting. Please feel free to suggest some mathematical descriptions.

Also thanks for the suggestion of tools. They look very helpful, but I really hope to understand some mathematics behind it.

  • 3
    $\begingroup$ Are you familiar with linkage disequilibrium and Hardy-Weinberg equilibrium? Looking up review papers on those might give more insights. $\endgroup$ – gringer Sep 1 '17 at 5:43
  • $\begingroup$ Re the edit: given the very low mutation rates in eukaryotes, mutation effect on HWE in SNPs should be negligible - or at least, much lower than the effect of genotyping errors. $\endgroup$ – juod Sep 2 '17 at 9:00

Simulating genotypes with realistic correlation structures is indeed not so simple, and there's quite a few papers dedicated entirely to that (e.g. https://bmcgenet.biomedcentral.com/articles/10.1186/s12863-015-0173-4). Also, DEPICT (https://data.broadinstitute.org/mpg/depict/index.html) comes with a number of simulated GWASs to generate the nulls, so that's a simple way to obtain some ready-made fake data.

On the other hand, indepedent SNPs should behave as draws from $Binomial(2, MAF)$. Note that the standard QC procedure of filtering out genotypes that deviate from Hardy-Weinberg equilibrium is just a goodness-of-fit test against $Binomial(2, \hat{MAF})$. If you employ this filter, some LD-pruning, and still get different results, consider posting more details about the discrepancies you see - that would be indeed unexpected.

| improve this answer | |
  • $\begingroup$ Thanks. I agree that independent SNPs behave as draws from Binomial. I think the discrepancies I observe is mainly on a population level. $\endgroup$ – Haohan Wang Sep 1 '17 at 15:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.