This paper presents WU-seq, a novel methodology for effectively analyzing the massive amounts of genome sequence data generated by the advancement of next-generation sequencing technology. Existing statistical methods suffer from significant limitations in analytical power due to the low frequency of rare variants and the high dimensionality of the data. Based on the nonparametric U statistic, WU-seq can be applied to a variety of phenotypes without making assumptions about disease models or phenotypic distributions. Through simulation and empirical studies, we demonstrate that WU-seq outperforms the existing SKAT method, especially when phenotypes follow long-tailed distributions, and performs similarly to SKAT even when the assumptions are met. Finally, we apply WU-seq to the Dallas Heart Study (DHS) data and discover an association between ANGPTL 4 and very low-density lipoprotein cholesterol.
