polyAdapt : characterizing polygenic adaptive architectures in the presence of strong linkage disequilibrium

Characterizing the genetic architecture of adaptation remains a central challenge in population genetics, particularly when multiple loci contribute to selected phenotypes. Two-genotype experimental evolution studies offer a powerful framework for this purpose, yet existing methods lack the capacity for genome-wide inference of selection targets and their coefficients under linkage. Here, we introduce polyAdapt, a novel iterative algorithm that identifies selected haplotype blocks and estimates the associated selection coefficients from Pool-Seq data in two-genotype experiments. Rather than exploring all possible combinations of selection targets simultaneously, polyAdapt sequentially incorporates targets of decreasing effect, accounting for linked selection at each step and optimizing both selection coefficients and effective population size through comparison of empirical and simulated replicate allele frequency trajectories. Using simulated data sets, we demonstrate that polyAdapt accurately recovers selection targets and coefficients for oligogenic architectures (5 and 13 targets) and provides a lower bound on the number of contributing loci for polygenic architectures (50 targets). Applied to yeast experimental evolution data, polyAdapt infers a highly polygenic architecture with at least 20 selected haplotype blocks per chromosome, consistent with independent estimates from the literature. polyAdapt thus represents a flexible and powerful tool for dissecting the genetic architecture of adaptation in experimental evolution studies.