Abstract
Accurate inference of genomic ancestry is critically important in human genetics, epidemiology, and related fields. Geneticists today have access to multiple heterogeneous population-based datasets from studies collected under different protocols. Therefore, joint analyses of these datasets require robust and consistent inference of ancestry, where a common strategy is to yield an ancestry space generated by a reference dataset. However, such a strategy is sensitive to batch artefacts introduced by different protocols. In this work, we propose a novel robust genome-wide ancestry inference method; referred to as SUGIBS, based on an unnormalized genomic (UG) relationship matrix whose spectral (S) decomposition is generalized by an Identity-by-State (IBS) similarity degree matrix. SUGIBS robustly constructs an ancestry space from a single reference dataset, and provides a robust projection of new samples, from different studies. In experiments and simulations, we show that, SUGIBS is robust against individual outliers and batch artifacts introduced by different genotyping protocols. The performance of SUGIBS is equivalent to the widely used principal component analysis (PCA) on normalized genotype data in revealing the underlying structure of an admixed population and in adjusting for false positive findings in a case-control admixed GWAS. We applied SUGIBS on the 1000 Genome project, as a reference, in combination with a large heterogeneous dataset containing auxiliary 3D facial images, to predict population stratified average or ancestry faces. In addition, we projected eight ancient DNA profiles into the 1000 Genome ancestry space and reconstructed their ancestry face. Based on the visually strong and recognizable human facial phenotype, comprehensive facial illustrations of the populations embedded in the 1000 Genome project are provided. Furthermore, ancestry facial imaging has important applications in personalized and precision medicine along with forensic and archeological DNA phenotyping.
Author Summary Estimates of individual-level genomic ancestry are routinely used in human genetics, epidemiology, and related fields. The analysis of population structure and genomic ancestry can yield significant insights in terms of modern and ancient population dynamics, allowing us to address questions regarding the timing of the admixture events, and the numbers and identities of the parental source populations. Unrecognized or cryptic population structure is also an important confounder to correct for in genome-wide association studies (GWAS). However, to date, it remains challenging to work with heterogeneous datasets from multiple studies collected by different laboratories with diverse genotyping and imputation protocols. This work presents a new approach and an accompanying open-source software toolbox that facilitates a robust integrative analysis for population structure and genomic ancestry estimates for heterogeneous datasets. Given that visually evident and easily recognizable patterns of human facial characteristics covary with genomic ancestry, we can generate predicted ancestry faces on both the population and individual levels as we illustrate for the 26 1000 Genome populations and for eight eminent ancient-DNA profiles, respectively.
