autosomal ancestry inference

Question

[J.F.Jiang]

Hi all,

I am very interested in ancestry inference.

1. autosomal ancestry I've followed the tutorial from https://github.com/pcgoddard/Burchardlab_Tutorials/wiki/ADMIXTURE#popfile.

However, decision on cluster information K is quite confusing.

For example, there are already 12 populations among my data. How can I train these data so that I can fully separate the populations.

1. Y haplogroup inference yhaplo https://github.com/23andMe/yhaplo/ offered by 23andme can do this, however, the ISOGG in this repo is too old. And the author did not update them more. So I am wondering if there is any other similar tool that can predict the haplogroup as well as building the ISOGG tree myself.

Junfeng

[prasundutta87]

Since you are already using admixture software, check out its manual: http://www.genetics.ucla.edu/software/admixture/admixture-manual.pdf

I am not sure why the link is not opening, probably some updation is going in the server where it is stored, but, in the document, an explanation has been provided on how to select cluster information K through estimation of cross-validation error.

Answer 1

[Kevin Blighe]

There are many ways to do it. You can identify haplotype-tagging SNPs via linkage disequilibrium (LD) or you can find another way, e.g., via principal components analysis. Here, I relate to 3 such methods:

• A: How to predict individual ethnicity information by using hapmap data (PCA-based)
• Produce PCA bi-plot for 1000 Genomes Phase III in VCF format (LD based)
• Haplotype Classification Using Copy Number Variation and Principal Components Analysis (haplotype-tagging CNV)

In the first case, the sensitivity/specificity of the model trained on 1000 Genomes data is >98%. I use this model to predict the ethnicity of unknown samples that I receive.