LEI: A Novel Allele Frequency-Based Feature Selection Method for Multi-ancestry Admixed Populations.

Area of Science:

• Genomics
• Population Genetics
• Statistical Genetics

Background:

• Next-generation sequencing enables large-scale genotyping for genomic characterization.
• Feature selection is crucial for analyzing high-dimensional genomic data in evolutionary biology and human genetics.
• Existing methods for ancestry proportion estimation are limited for three-way admixed populations, such as Latino populations.

Purpose of the Study:

• To introduce a novel likelihood-based feature selection method, the Lancaster Estimator of Independence (LEI).
• To utilize allele frequency data for prioritizing informative genetic markers for ancestry proportion estimation.
• To address the limitations of current statistical tools in analyzing three-way admixed populations.

Main Methods:

• Developed LEI, a likelihood-based feature selection method using allele frequency information.
• Leveraged summary statistics from allele frequency data, avoiding individual-level genotype data requirements.
• Compared LEI's allele-frequency based approach with genotype-based methods in three-way admixed population simulations.

Main Results:

• LEI effectively prioritizes informative genetic markers for ancestry proportion estimation.
• Ancestry estimates using LEI-selected features were comparable to those from genotype-based methods.
• The LEI approach minimizes computational demands and time for ancestry inference in admixed individuals.

Conclusions:

• LEI offers an efficient and accessible tool for feature selection in population genetics.
• The method is particularly valuable for analyzing complex admixed populations, including Latino populations.
• Provided R code facilitates the application of LEI in admixture mapping studies.