Using a Machine Learning Algorithm to Identify Key Loci Involved in Ecological Speciation in Chinook Salmon (Oncorhynchus tshawytscha) in the Columbia River Basin and Puget Sound, USA

The conservation of biodiversity at the genetic level in fish populations will increasingly rely on molecular markers that are linked to fitness traits. Current efforts are focused on characterizing these markers through a variety of approaches that link genotypes with phenotypes. High throughput sequencing technologies have resulted in datasets where the number of loci genotyped is much greater than the number of individuals sampled. Such large datasets present statistical challenges associated with multiple testing. Here, we present efforts to characterize loci that may have been involved in a key trait, run timing, which appears to have undergone parallel evolution in populations of Chinook salmon in the Columbia River basin in the Pacific Northwest, USA. We characterized over 9000 Restriction-site Associated DNA loci, many of which are mapped, across eleven populations in the region.  We used analytical approaches that integrated information across multiple outlier tests partitioned between evolutionary lineages. We also used a novel statistical technique that identified predictors of traits in situations where there are more variables than observations.  This integrative approach revealed neutral evolution and high resolution population assignment. We also detected loci that accurately predicted run timing and provided evidence for parallel evolution in this trait.