LoGicAl: Local ancestry and genotype calling uncertainty-aware ancestry-specific allele frequency estimation from admixed samples

Abstract

Motivation: In admixed groups, it is of interest to estimate allele frequencies of their ancestral contributing populations. These ancestry-specific allele frequencies inform genetic drivers of disease etiologies, facilitate genome-wide association study interpretations, enhance polygenic risk prediction and portability, and provide insights into demographic history. Their estimation leverages inferred locus-specific ancestry background, i.e., local ancestry, generated by tools like RFMix. However, existing estimation methods lose accuracy and incur biases by failing to model uncertainty from upstream local ancestry inference and genotype calling. Results: Here, we introduce LoGicAl, a novel likelihood-based method for estimating ancestry-specific allele frequencies from admixed samples, simultaneously accommodating uncertainty from ancestry calling, genotyping, and statistical phasing. We demonstrate that modeling these uncertainties substantially reduces estimation errors, resulting in superior accuracy for both sequence-based and array-based genotyping with different levels of local ancestry inference quality. By integrating an accelerated fixed-point algorithm, LoGicAl achieves high scalability and enhanced computational efficiency compared with existing approaches. Applying LoGicAl to admixed cohorts in the 1000 Genomes Project, we illustrate the benefits of local-ancestry-based allele frequency estimates. Together, LoGicAl contributes to genomic analyses of admixed samples by providing precise and rapid ancestry-specific allele frequency estimates, and to constructing the spatial landscape and dynamics of genetic variations in admixed populations at a finer scale. Availability: The LoGicAl software is available at https://github.com/student-jw/LoGicAl.