An Extension to and Application of the Multifactor Dimensionality Reduction Pedigree Disequilibrium Test

Abstract

As the field of genetics explores beyond mapping single-site disease susceptibility loci, epistasis between genes is being considered in disease models. These hypotheses present new problems for investigators as they search through ever more complex data structures. The dimensionality and size of a search space, the types and strengths of disease associations in data, and the quality of inference allowed given a result are all challenges when testing for putative epistatic disease models. Method development to analyze family data for epistatic interactions is in a preliminary stage. The multifactor dimensionality reduction pedigree disequilibrium test (MDR-PDT) is one technique for assessing epistatic models in family data. The objective of this proposal is to refine, test, and apply this method to real data. MDR-PDT is a method that implements the genoPDT statistic within the framework of the MDR algorithm. We hypothesize that at the conclusion of my aims, the MDR-PDT algorithm’s utility and power will be improved. In the following dissertation we developed a cross validation algorithm for pedigree data, and an omnibus model selection method. We also implemented an extension to MDR-PDT that includes a likelihood ratio test for the statistical significance of an interaction found by the MDR-PDT search using logistic regression. Finally, MDR-PDT was applied to Alzheimer’s candidate gene datasets and revealed a multilocus model involving several genes that are functional candidates.