Estimation of individual admixture: analytical and study design considerations

Area of Science:

• Population Genetics
• Genomic Ancestry Analysis
• Epidemiological Research

Background:

• Admixed individuals, common in the US (e.g., African-Americans, Hispanics), have genomes mixing different ancestries.
• Understanding individual admixture proportions (IA) is vital for geneticists and epidemiologists in case-control association studies.

Purpose of the Study:

• To present an extended frequentist maximum likelihood (ML) method for estimating individual admixture.
• To address uncertainty in ancestral allele frequencies during IA estimation.

Main Methods:

• Extension of a previously described frequentist ML approach.
• Comparison with partial likelihood methods and Bayesian Markov Chain Monte Carlo (MCMC) methods.
• Simulations to evaluate robustness, efficiency (mean squared error), and computational time.

Main Results:

• The full ML method shows improved robustness over existing partial ML approaches.
• The frequentist estimator achieves comparable efficiency to Bayesian methods.
• The ML approach requires significantly less computational time than Bayesian methods, enabling extensive analyses.

Conclusions:

• The developed frequentist ML method provides a robust and computationally efficient tool for individual admixture estimation.
• Accurate IA estimation necessitates the inclusion of ancestral populations or their surrogates in the analysis.