Search:

Cancer Control Research

Abstract

DESCRIPTION (provided by applicant): The success of genetic dissection of complex diseases may greatly benefit from judicious exploration of joint gene effects or gene-gene interactions, which, in turn, critically depends on the power of statistical tools. Standard regression models are convenient for assessing main effects and low-order interactions but not for exploring complex higher- order gene-gene interactions. Tree-based methodology is an attractive alternative for disentangling possible interactions, but it has difficulty in modeling additive main effects. We propose a new class of semiparametric regression models, termed partially linear tree-based regression (PLTR) models, that has the advantages of both generalized linear regression and tree models. A PLTR model quantifies joint effects of genes by a combination of linear main effects and a nonparametric tree-structure. In particular, it permits the use of linear terms for confounder variables and a tree structure for the joint effect of interest. The first specific aim is to assess an iterative algorithm to fit the PLTR model and to develop methods for identifying and testing the significance of an optimal-pruned tree nested within the tree resultant from the fitting algorithm. The second specific aim is to apply the developed method in analyses of several cancer genetic epidemiology studies to explore joint gene-gene or gene-environment interactions.