Genome Scans for Selection and Introgression based on k-nearest Neighbor Techniques

Abstract

In recent years, genome-scan methods have been extensively used to detect local signatures of selection and introgression. Here, we introduce versatile genome-scan methods based on non-parametric k-nearest neighbors (kNN) techniques, while incorporating pairwise Fixation Index (F ST ) estimates and pairwise nucleotide differences (d xy ) as features. Simulations were performed for both positive directional selection and introgression, with varying parameters, such as recombination rates, population background histories, the proportion of introgression, and the time of gene flow. We find that kNN-based methods perform remarkably well while yielding stable results almost over the entire range of k. Furthermore, the weighted-kNN algorithm for detecting directional selection, and the INFLO (Influenced Outlierness) algorithm for detecting introgression, outperform recently published methods. We provide a GitHub repository (pievos101/kNN-Genome-Scans) containing R source code to demonstrate how to apply the proposed methods to real-world genomic data using the population genomics R-package PopGenome.