In population genomics studies, accounting for the neutral covariance structure across population allele frequencies is critical to improve the robustness of genome-wide scan approaches. Elaborating on the BAYENV model, this study investigates several modeling extensions i) to improve the estimation accuracy of the population covariance matrix and all the related measures; ii) to identify significantly overly differentiated SNPs based on a calibration procedure of the XtX statistics; and iii) to consider alternative covariate models for analyses of association with population-specific covariables. In particular, the auxiliary variable model allows to deal with multiple testing issues and, providing the relative marker positions are available, to capture some Linkage Disequilibrium information. A comprehensive simulation study was carried out to evaluate the performances of these different models. Also, when compared in terms of power, robustness and computational efficiency, to five other state-of-the-art genome scan methods (BAYENV2, BAYSCENV, BAYSCAN, FLK and LFMM) the proposed approaches proved highly effective. For illustration purpose, genotyping data on 18 French cattle breeds were analyzed leading to the identification of thirteen strong signatures of selection. Among these, four (surrounding the KITLG, KIT, EDN3 and ALB genes) contained SNPs strongly associated with the piebald coloration pattern while a fifth (surrounding PLAG1) could be associated to morphological differences across the populations. Finally, analysis of Pool-Seq data from 12 populations of Littorina saxatilis living in two different ecotypes illustrates how the proposed framework might help addressing relevant ecological issue in non-model species. Overall, the proposed methods define a robust Bayesian framework to characterize adaptive genetic differentiation across populations. The BAYPASS program implementing the different models is available at http://www1.montpellier.inra.fr/CBGP/software/baypass/.
Stratification Methods for an Auxiliary Variable Model-Based Allocation under a Superpopulation Model
