Genetic Basis of Adult Migration-Timing within a Population of Steelhead (Oncorhynchus mykiss)

Migration traits are presumed to be complex and to involve interaction among multiple genes of minor effect, therefore univariate analyses may lack the power to detect associations.  Further, analyses of species that lack a wealth of genomic resources (e.g. whole genome sequence, high density linkage maps, or >100 k markers for genotyping) may present difficulties in elucidating the key genomic regions underpinning particular traits.  Therefore we employed both univariate analyses as well as a random-forest (RF) machine learning algorithm to conduct association mapping of ~15k SNPs for adult migration-timing phenotype in Steelhead trout (Oncorhynchus mykiss).  A univariate mixed linear model found 3 SNPs to be significantly associated with migration-timing.  The same 3 SNPs were ranked high in importance values based on RF and explained 46% of trait variation (7% residual).  However, RF identified 44 minor SNPs were required to reach a maximum ~70% explained trait variation (46% residual).  These candidate SNPs may provide the ability to predict the adult migration-timing of Steelhead trout and facilitate conservation management of this species.  This study also demonstrates how beneficial a RF approach may be for identifying SNPs in minor effect genes of complex traits that may otherwise escape detection in univariate association analyses.