Human ancestry inference at scale, from genomic data

Abstract

Using an alignment-free single nucleotide variant prediction framework that leverages integrated variant call sets from the 1000 Genomes Project, we demonstrate accurate ancestry inference predictions on over 600 human genome sequencing datasets, including complete genomes, draft assemblies, and >280 independently-generated datasets. The method presented, ntRoot, infers super-population ancestry along an input human genome in 1h15m or less on 30X sequencing data, and will be an enabling technology for cohort studies.

A comprehensive tandem repeat catalog of the human genome

“…Information on the population origins of our study samples was accessible from the 1000 Genomes study, which informs that our samples originate from five superpopulations: African (AFR, n=87), European (EUR, n=24), East Asian (EAS, n=55), admixed American (AMR, n=56), and South Asian (SAS, n=50) ( Supplementary table 6 ). While previous studies have used single nucleotide variants (SNVs) as the vehicle for genetic ancestry inferences 50 , 51 , we explored whether polymorphic TR genotypes alone can be used to segregate samples to their origin populations. All TRs that have all 544 haplotypes genotyped and a standard deviation of at least two repeat copies (n=782) were subjected to a principal component analysis (PCA).…”