On the sensitivity to hyperparameters of a Bayesian method to infer selection signatures (Simulated Data)

Genotypes of 1247 individuals distributed in three subpopulations for 1000 biallelic molecularmarkers were simulated using a forward in time approximation as implemented in the QMSIMsoftware (Sargolzaei and Schenkel, 2013). A historical population comprised of 1000 generations ofsize 5000 was created and three random samples were taken to generate three subpopulations (S1,S2 and S3) which had different mating systems (assortative for S1 and S2 and random for S3),selection criteria (high breeding values for S1, low breeding values for S2, low phenotypes for S3)and numbers of generations (25 for S1, 20 for S2, and 4 for S3), sample sizes were 506, 441 and300, respectively; these individuals were chosen from the last generations of each subpopulation.The heritability of the trait was 0.5 and its phenotypic variance was 100, the number ofchromosomes was 10, all having a length of 10 centimorgan with 100 markers per chromosome.This procedure was replicated 10 times. The aim of this simulation strategy was creating differencesin allele frequencies across subpopulations.