摘要: In recent years,genome-wide association study (GWAS) has been used as the default approach for dissecting the genetic architecture of complex traits of human,plants and animals.With convenience of next generation sequencing techno... 展开 In recent years,genome-wide association study (GWAS) has been used as the default approach for dissecting the genetic architecture of complex traits of human,plants and animals.With convenience of next generation sequencing technology,a large amount of high dense single nucleotide polymorphism (SNP) data have been generated for different organisms. GWAS is computationally challenging due to large amount of SNP data, whereas a large sample size is required for identifying the underlying genetic variants of complex traits. Again,false positive due to several confounding factors is also major concern in GWAS. Researchers are mostly paying attention to detect loci by testing their marginal effects.However,unbiased estimation of parameters is also important for estimating the total genetic breeding value due to detectable causal variants that could help in genetic breeding and personalize medicine.Testing marginal effects of loci could give biased estimates of the parameters and might be cause of misleading results.Dominance,epistasis and environmental specific genetic effects might have influences on complex traits that are being ignored in most of the GWAS. Ignoring dominance and epistasis interactions is one of the major causes of missing heritability of GWAS. In this study,large amount of heritability was estimated due to dominance and epistasis interaction for total cholesterol of human multi-ethnic study of atherosclerosis (MESA) population and two leaf traits of maize NAM population. The four different approaches were compared:full model approach(including additive,dominance,epistasis and ethnic interactions),multi-loci additive model approach and two popular single locus approach implemented in PLINK and GCTA by using real data of total cholesterol and simulated data.Analyses of cholesterol data and simulation study showed full model approach outperforms over additive model approaches in terms of unbiased estimation of genetic parameters and detection powers of underlying genetic variants of complex traits.Ignoring important genetic variants could reduce detection power of loci,parameter estimation could be biased upward or downward,and mean sum of square errors(MSE)of genetic parameters could be increased. In this study,genetic architecture of total cholesterol of MESA population and maize leaf traits were dissected that provide new insight about the traits.Analysis with total cholesterol data identified3previously reported gene and several new genes.Bioinformatics analysis of the candidate gene showed some newly identified genes have relationship with cholesterol related chemical or cholesterol related diseases. Leaf traits of maize NAM population were analyzed using full model and multi-loci additive model.Estimated heritability using full model approach was larger than the estimated heritability using additive model approach for the leaf traits.The estimated heritability for upper leaf angle was mostly for additive and additive-by-additive epistasis effects,whereas for leaf length and leaf width was mostly for dominance related epistasis interactions,indicating that the genetic architecture of maize upper leaf angle is different from other two traits.Only few pleotropic loci were identified that refers different sets of genes are controlling the leaf traits.Superior line,superior hybrid were predicted by utilizing association analysis results.By comparing genetic breeding value estimated from additive model and full model,more scope of further trait manipulation was observed by accounting heterozygous genotypes.This study describes the utility of full model approach for GWAS of complex trait analysis. 收起