AbstractAssociation mapping (AM) methods are used in genome-wide association (GWA) studies to test for statistically significant associations between genotypic and phenotypic data. The genotypic and phenotypic data share common evolutionary origins – namely, the evolutionary history of sampled organisms – introducing covariance which must be distinguished from the covariance due to biological function that is of primary interest in GWA studies. A variety of methods have been introduced to perform AM while accounting for sample relatedness. However, the state of the art predominantly utilizes the simplifying assumption that sample relatedness is effectively fixed across the genome. In contrast, population genetic theory and empirical studies have shown that sample relatedness can vary greatly across different loci within a genome; this phenomena – referred to as local genealogical variation – is commonly encountered in many genomic datasets. New AM methods are needed to better account for local variation in sample relatedness within genomes.We address this gap by introducing Coal-Miner, a new statistical AM method. The Coal-Miner algorithm takes the form of a methodological pipeline. The initial stages of Coal-Miner seek to detect candidate loci, or loci which contain putatively causal markers. Subsequent stages of Coal-Miner perform test for association using a linear mixed model with multiple effects which account for sample relatedness locally within candidate loci and globally across the entire genome.Using synthetic and empirical datasets, we compare the statistical power and type I error control of Coal-Miner against state-of-theart AM methods. The simulation conditions reflect a variety of genomic architectures for complex traits and incorporate a range of evolutionary scenarios, each with different evolutionary processes that can generate local genealogical variation. The empirical benchmarks include a large-scale dataset that appeared in a recent high-profile publication. Across the datasets in our study, we find that Coal-Miner consistently offers comparable or typically better statistical power and type I error control compared to the state-of-art methods.CCS CONCEPTSApplied computing → Computational genomics; Computational biology; Molecular sequence analysis; Molecular evolution; Computational genomics; Systems biology; Bioinformatics; Population genetics;ACM Reference formatHussein A. Hejase, Natalie Vande Pol, Gregory M. Bonito, Patrick P. Edger, and Kevin J. Liu. 2017. Coal-Miner: a coalescent-based method for GWA studies of quantitative traits with complex evolutionary origins. In Proceedings of ACM BCB, Boston, MA, 2017 (BCB), 10 pages. DOI: 10.475/123 4
Empirical Comparison of Tests for One-Factor ANOVA Under Heterogeneity and Non-Normality: A Monte Carlo Study
