A Nonparametric Method To Test For Associations Between Rare Variants And Multivariate Phenotypes

The emergence of the next generation high-throughput DNA sequencingtechnology provides us a huge number of genetic data of rare variants. It is a currenthot problem how to define rare variants associated with human complex disease fromdata sets. Recently，people have already developed several methods to test forassociation between rare variants and human complex disease, and the most of theexisting methods test multiple locus and one trait. However, some complex diseasesespecially mental illness and behavioral disorders, the traits can be recorded indifferent scales, such as binary, ordinal and quantitative. Furthermore, comprehensivestudies have demonstrated that jointly testing correlated traits is more powerful thantesting a single trait at a time. Therefore, this article proposes a nonparametricassociation analysis method based on the generalized Kendall’s τ method to test thecorrelation between rare variants and multivariate phenotype, where trait value can bebinary, ordinal, quantitative and random mixture of them. We introduce a new kernelfunction, which is used to measure difference of phenotype and genotype of differentindividuals; take advantage of statistic to construct test statistics and derive theasymptotical distribution of the statistics. At last, through simulation study based onreal data sets, we compare our method and other methods, and the results show thatour proposed method is more power and robust in testing multivariate phenotypes,especially for ordinal multivariate phenotypes.