Jointly Analyze Multiple Phenotypes Under Model Uncertainty Using Clustering Linear Combination Approach In GWAS

Genome-wide association studies is a very effective research tool for identifying potential genetic variants of various complex diseases.New evidence shows that genetic variation can affect multiple phenotypes,especially in complex human diseases.The combined analysis of multiple phenotypes may not only increase the statistical power to detect genetic variants associated with complex diseases,but may also provide new insights into the etiology of diseases.Therefore,people increasingly hope to conduct joint analysis of multiple phenotypes for genome-wide association studies.In recent years,many statistical methods have been developed for the joint analysis of multiple phenotypes.Among them,the hierarchical clustering linear combination(HCLC)method has many advantages compared with other existing methods.The HCLC method can obtain and use the natural grouping information of the phenotype from the hierarchical clustering.It is easy to implement and has high computational efficiency.It is a powerful test statistics method.Existing methods are to construct test statistics on the premise of a known genetic model.However,in actual research,the genetic model is usually unknown.Since the uncertainty of the genetic model is not considered,these methods may lead to unstable performance.When the genetic model is uncertain,a robust test statistic is a good choice.The MAX3 test is a commonly used robust test statistic.This article first introduces the basic knowledge of human genetics and test statistics commonly used in genome-wide association analysis.Next,we introduced the principles and methods commonly used in joint analysis of genetic associations of multiple phenotypes.Then,considering the uncertainty of the genetic model,this article proposes a MAX3-based HCLC test method(MHCLC,that is,the absolute value of the HCLC statistics calculated under the dominant,additive,and recessive genetic model takes the maximum value)to jointly analyze multiple phenotypic genetic association issues.The statistical theory properties of the MHCLC method are discussed,and the first type error rate and efficacy of the method are evaluated through simulation studies.Numerical results show that,when the considered genetic model is uncertain,the MHCLC method can effectively control the first type of error rate and has good statistical robustness.