search

slowkoni / rfmix

RFMIX - Local Ancestry and Admixture Inference Version 2
79 stars 25 forks source link

Unexpected global diploid results. #32

Open mfknie opened 3 years ago

mfknie commented 3 years ago

I am currently trying to run RFmix on an African American population (with ~5700 individuals). I am using two reference panels from 1000G phase 3, using the labeled AFR and EUR individuals (with 661 for the former, and 503 for the latter). As such, we would expect the global ancestry proportions to be about 80% AFR and 20% EUR.

After running RFmix (using 5 em results), I inspected the global diploid estimates in rfmix.Q, and found that the ancestry proportions were the opposite of what I expected; with an average around 0.17 for the AFR population, and 0.83 for the EUR population (taking a weighted average across all chromosomes). I double checked the population labels in the sample map file, and found that they were correct. I tried switching the labels around, and got the expected results. Are there any possible explanations for this?

EfraMP commented 6 months ago

Why did you choose that specific number of iterations?