Rare variants play a critical role in explaining the genetic contribution to complex diseases by accounting for disease risk and trait variability. Building association tests of rare variants is challenging using the existing tools that merely compare their frequencies across different groups. In this talk, we will introduce a powerful new statistical method that incorporates spatial locations of variants, allows incorporation of previous gene ontology information, and appropriately characterizes uncertainty in inferences. We will build a multiresolution cluster detection algorithm using a binary-tree-based nonparametric Bayes test and show how it leads to robustness, adaptability to the underlying disease architecture, and biologically relevant segmentation of the genome. We will apply our method on 240 cases of patients with primary immune deficiency to identify patterns of genetic variation underlying the disease, compared to over 650,000 controls. We will end with extension of this model across a large range of disease models, while maintaining scalability, and incorporation of important covariates and adjustment for population stratification.