Genome scans for detecting footprints of local adaptation using a Bayesian factor model

Area of Science:

• Population genomics
• Evolutionary biology
• Bioinformatics

Background:

• Identifying genomic regions under local adaptation is crucial in population genomics.
• Traditional methods like FST rely on predefined population structures, which can be limiting.
• Existing approaches may inaccurately identify outlier loci due to assumptions about population structure.

Purpose of the Study:

• To develop a flexible, individual-based approach for detecting local adaptation.
• To infer population structure and identify outlier loci simultaneously.
• To improve the accuracy and reduce false discoveries in selection scans.

Main Methods:

• Implementation of a hierarchical Bayesian factor model.
• Utilizing latent variables (factors) to capture complex population structures (e.g., clustering, isolation-by-distance).
• Identifying outlier loci by their atypical relationship to inferred population structure.

Main Results:

• The Bayesian factor model significantly reduces the false discovery rate compared to FST and BayeScan.
• Demonstrated effectiveness on large datasets, including Human Genome Diversity Project single nucleotide polymorphisms.
• Successfully infers population structure and detects loci under selection simultaneously.

Conclusions:

• The Bayesian factor model offers a more robust and flexible alternative for detecting local adaptation.
• This individual-based approach enhances the accuracy of selection scans in population genomics.
• The open-source PCAdapt software provides a powerful tool for researchers in the field.