CLEAR: Composition of Likelihoods for Evolve And Resequence Experiments

Experimental evolution (EE) studies are powerful tools for observing molecular evolution \"in-action\" from populations sampled in controlled and natural environments. The advent of next generation sequencing technologies has made whole-genome and whole-population sampling possible, even for eukaryotic organisms with large genomes, and allowed us to locate the genes and variants responsible for genetic adaptation. While many computational tests have been developed for detecting regions under selection, they are mainly designed for static (single time) data, and work best when the favored allele is close to fixation.\n\nEE studies provide samples over multiple time points, underscoring the need for tools that can exploit the data. At the same time, EE studies are constrained by the limited time span since onset of selection, depending upon the generation time for the organism. This constraint impedes adaptation studies, as the population can only be evolve-and-resequenced for a small number of generations relative to the fixation time of the favored allele. Moreover, coverage in pool-sequenced experiments varies across replicates and time points, leading to heterogeneous ascertainment bias in measuring population allele frequency across different measurements.\n\nIn this article, we directly address these issues while developing tools for identifying selective sweep in pool-sequenced EE of sexual organisms and propose Composition of Likelihoods for Evolve-And-Resequence experiments (CO_SCPLOWLEARC_SCPLOW). Extensive simulations show that CO_SCPLOWLEARC_SCPLOW achieves higher power in detecting and localizing selection over a wide range of parameters. In contrast to existing methods, the CO_SCPLOWLEARC_SCPLOW statistics are robust to variation of coverage. CO_SCPLOWLEARC_SCPLOW also provides robust estimates of model parameters, including selection strength and overdominance, as byproduct of the statistical test, while being orders of magnitude faster than existing methods. Finally, we apply the CO_SCPLOWLEARC_SCPLOW statistic to data from a study of D. melanogaster adaptation to alternating temperatures and discover enrichment of genes related to \"response to heat\" and \"cold acclimation\".