Chromosomal Admixture Painting

You have likely seen the 23andme ancestry painting and probably Doug McDonald's chromosome painting too. This is not quite the same.

Instead of looking at ancestry of segments, I am looking at ancestry of whole chromosomes. My curiosity for this analysis derives from my lovingly homozygous chromosome 9 and a-little-African chromosome 8.

Basically, this is the same Admixture analysis with Reference I dataset and a batch of Harappa project participants, except that instead of using all 22 chromosomes, I ran admixture separately on each chromosome.

Since each chromosome's data was separately processed, the ancestral components inferred for each chromosome are not exactly the same. In practice, at K=4 ancestral components, they stayed reasonably constant, but the errors are larger than the overall autosomal admixture analysis. So you should be wary of assigning significance to minor changes in percentages from chromosome to chromosome. As a rule of thumb, a difference of greater than 5% from the autosomal should be required for you to give it some thought.

I have run the same admixture analysis with K=6 for this group. Once I have analyzed that, I'll write about it. If you guys think it's something worth doing, then I can run the same analysis for the latter batches. Otherwise, we can look directly into proper chromosomal segment painting.


  1. This chart, I can see.

  2. Excellent work Zack, just excellent. Greatly appreciate the effort.

  3. Wow, did you actually do 220 ADMIXTURE runs for this? Impressive.