A High-efficiency Method For Multiple Traits Mixed Model Association Analyses

Owing to their ability to comprehensively account for confounding factors,linear mixed model(LMM)has become a powerful tool in genome-wide association studies(GWAS)for multiple correlated phenotypes.Multivariate Linear mixed model(mv LMM)association analyses can detect not only pleiotropic QTNs but also no-pleiotropic QTNs that control only one of multiple traits.They result in gains of statistical powers to detect QTNs over standard univariate analysis,and biologically,can explain how genetic relationships happen among multiple traits by using pleiotropic QTNs and no-pleiotropic QTNs with high linkage disequilibrium.Based on genomic variance components,we simplified multivariate mixed model association analysis to multiple univariate ones by using canonical transformation,and then individually implemented univariate association tests in the Fa ST-LMM.which enlarged numbers of the analysed phenotypes.With canonical transformation back to the original scale,the association results would be biologically interpretable.Further,we rapidly estimated genomic covariance matrices with multivariate GEMMA and optimized the polygenic heritabilities for only the makers that had large effects or higher significance levels,greatly improving computing efficiency for multiple univariate association tests.Beyond one test at once,joint association analysis significantly increased statistical powers to detect quantitative trait nucleotide candidates.A userfriendly mvRunKing software was developed to efficiently implement multivariate mixed model association analyses.We demonstrate statistical utility of mvRunKing by computer simulations.Continuous quantitative traits are simulated based on genomic datasets in human and maize,respectively.This concluded:(1)mvRunKing before optimisation could detect QTNs in almost the same statistical power as GEMMA with Wald tests,when genetic correlation was far greater than the residual correlations between traits in the population with lowly complex structure.As long as the residual correlation is not too large,PCA for multiple traits also could yield the same outcomes as mvRunKing after optimisation.Additionally,with the consideration of possible linkage disequilibrium among candidate markers,joint-association analysis for multiple QTN candidates obtained with multiple testing allowed mvRunKing to improve the statistical power in detecting QTNs.(2)As the number of analysed traits increased,GEMMA dramatically increased the computational times,while mvRunKing little without limitation of memory.mvRunKing reduced the computational time by several times to hundredfold of times compared with GEMMA in the cases of simulated and real phenotypes.Genome-wide association analysis of multiple growth traits in Nile tilapia(Oreochromis niloticus)proved that mvRunKing can guarantee the same statistical power as the GEMMA,and the operation time is several to hundreds of times faster than it.