Old vs. New Local Ancestry Inference in HCHS/SOL: A Comparative Study.

Area of Science:

• Population Genetics
• Genomic Ancestry Inference
• Statistical Genetics

Background:

• Hispanic/Latino populations exhibit genetic admixture, requiring accurate local ancestry inference for genetic association studies.
• Local ancestry estimation is crucial for admixture mapping, relying on reference panels and inference methods.
• The choice of inference method and reference panel significantly impacts the accuracy of local ancestry estimations.

Purpose of the Study:

• To evaluate and compare the performance of a new local ancestry inference method (FLARE) against a standard method (RFMix) using Hispanic/Latino data.
• To assess the impact of updated reference panels on local ancestry inference accuracy.
• To determine if differences in local ancestry inference affect admixture mapping results.

Main Methods:

• Utilized the Hispanic Community Health Study/Study of Latinos (HCHS/SOL) dataset comprising Hispanic/Latino individuals.
• Employed RFMix for "old" local ancestry inference and Fast Local Ancestry Estimation (FLARE) with an updated reference panel for "new" inference.
• Compared global and local ancestry correlations and evaluated admixture mapping-based associations between the two methods.

Main Results:

• Both RFMix and FLARE demonstrated high similarity in global and local ancestry correlations.
• FLARE-inferred datasets produced admixture mapping results consistent with those from RFMix.
• Lower correlations between old and new local ancestries were observed in specific genomic regions, often coinciding with ENCODE blacklist regions or gene clusters.

Conclusions:

• Fast Local Ancestry Estimation (FLARE) offers a reliable alternative for local ancestry inference in admixed populations like Hispanic/Latino individuals.
• The updated reference panel used by FLARE did not significantly alter admixture mapping outcomes compared to the established RFMix method.
• Discrepancies in local ancestry inference primarily occurred in complex genomic regions, suggesting potential limitations in current inference methods for such areas.