Research And Application Of Population-based Association Analysis For Gene Mapping

Genome-wide association studies can identify a large number of common variants associated with complex disease.However,these common variants can only account for a small fraction of disease heritability.Some evidences show that rare variants can contribute to most of the heritability of complex disease.Because of low minor allele frequency of rare variant,common variant association analysis is not optimal for detecting rare variant.Therefore,rare variant association analysis has become the focus of research in recent years.In addition,pleiotropy(the effect of one variant on multiple traits)is widespread in complex diseases.Joint analysis of multiple traits can improve statistical power to detect genetic variants with pleiotropy effects,and uncover the underlying genetic mechanism.Thus,more and more powerful methods of association analysis are needed to joint analysis of multiple traits.Based on population data,the associations between genetic variants and single trait or multiple traits are investigated in this thesis.Firstly,a method based on Fisher's combination of P-values is proposed to test the association between rare variants and a quantitative trait.Each rare variant is tested by the score test,and P-values of single-variant tests are obtained.According to the correlation between each rare variant and the quantitative trait,directions of effects of rare variants are distinguished reasonably.For different directions,P-values of rare variants with strong association signals are separately combined with a suited weight,which depends on the minor allele frequency and covariance between each rare variant and the trait.The simulation studies show that the approach is powerful,when the genetic region contains a large number of noncausal variants.Secondly,based on the reverse regression model,a method by adaptive combination of P-values is proposed to test association between rare variants and multiple traits.The reverse regression model is used to test each rare variant associated with multiple quantitative traits,and P-values of single variant tests are obtained.Based on the correlation between each rare variant and the first principal component of multiple traits,directions of effects of rare variants are distinguished reasonably.According to different directions,Pvalues of single-variant tests are adaptively combined with the weight,which depends on the minor allele frequency and covariance between each rare variant and the first principal component of multiple traits.The simulation studies show that when the genetic region contains a large number of noncausal variants,the approach is powerful;the approach is efficient in the presence of noise traits.Finally,a dimension reduction method is proposed to detect both rare and common variants associated with multiple traits.Association between each trait and genetic variants is tested,and test statistics of single trait are obtained.For the purpose of dimension reduction,these test statistics are used as weights to combine original traits,and the linear combination of original traits is obtained.Then association between the linear combination of traits and multiple variants is tested.However,this method loses power in the presence of noise traits.Based on this method,a method of excluding noise traits is proposed.The simulation studies and real data analysis show that the proposed method is feasible and efficient as the region-based method.