A novel framework for assessing the power of genomic animal models in wild non-model organisms

O_LIThe relative contributions of heritable versus non-heritable phenotypic variation determine how selection may shape evolutionary responses. While genome-wide association studies can provide some insight into large-effect loci underpinning phenotypes, they potentially miss much of the evolutionarily important variation in polygenic traits because many small-effect loci will contribute to heritable variation. C_LIO_LIGenomic animal models can estimate trait heritability from genomic data but are often difficult to apply to wild organisms. Sampling groups of close kin improves their power, yet this can be logistically challenging in species with large populations and high dispersal. Furthermore, many common approaches for power simulations rely on kin structure in the genomic relationship matrix, A, but such simulations on an unstructured A (comprised of unrelated individuals) are likely uninformative. C_LIO_LIWe developed a novel framework to generate simulated expectations from genomic animal models when applied to samples of unrelated wild organisms. A key innovation of this framework is the use of a phenotypes-from-genotypes approach that allows quantitative traits to be simulated from genotypes, independently of A. This provides clear utility where kin structure is absent. We illustrated the use and flexibility of our phenotypes-from-genotypes simulation approach to explore how the performance of genomic animal models can be influenced by different genetic architectures (clumped vs even genomic distribution of QTLs, and different QTL effect sizes), levels of heritability, and experimental design (sampled individuals and genetic markers). C_LIO_LIWe then applied our framework to a study of head shape traits in a wild marine intertidal fish, Bathygobius cocosensis (Bleeker 1854), for which we acquired SNP genotypes with RADseq. Although our working sample of 297 fish contained no close kin, our framework helped validate the power of genomic animal models, allowing us to confidently infer heritable variation in one head shape trait (estimated h2 = 0.17). C_LIO_LIOur study addresses key methodological gaps for using quantitative genomic approaches on wild organisms, offering a rare example of partitioned heritable and non-heritable contributions to phenotypic variation in a non-model wild marine fish. Our novel phenotypes-from-genotypes approach also provides a new method for simulating QTLs using observed population genetic data. C_LI