Computational Methods to Detect Epistasis

点击翻译按钮获取中文摘要

The completion of the draft sequence of the human genome and the continuing development of technologies to rapidly genotype single nucleotide polymorphisms (SNP) have led to great optimism that researchers will be able to determine the genetic mechanisms contributing to many common diseases. However, thus far there has been little success in finding genes to account for the large genetic components of common, complex diseases such as cancer, diabetes, depression, alcoholism, and asthma. These complex phenotypes are likely to be associated with gene interactions more complicated than simple additive or multiplicative models suggest. While great progress had been made on Mendelian diseases using single-locus models, the failure of these methods to determine the genetic contributors to many of the common diseases that are so costly to society underscores the need to develop new methodologies specifically aimed at detecting gene interactions. The goal of this proposal is to develop and evaluate novel methods for detecting gene-gene (epistatic) interactions and to apply them to pharmacogenetic data relating to drug metabolism. Specifically, the aims are: 1) to develop statistical methods to test for epistasis between candidate loci, focusing on methods to control the loss of power associated with correcting for multiple tests, 2) to evaluate the effectiveness of these detection and inference methods on simulated data, and 3) to apply and test these methods on real data from the Pharmacogenetics Research Network project headed by Dr. Howard McLeod at Washington University in St. Louis (U01 GM63340, Functional Polymorphism Analysis in Drug Pathways) and from warfarin metabolism studies headed by Dr. Brian Gage, also at Washington University (R01 HL71038 and R01 HL074724). Achieving these aims will require, in addition to mathematical skills, both statistical expertise and biological knowledge of genetics and metabolic pathways supplemented by knowledge of the laboratory techniques used to generate data for analysis. To this end, the proposed award will supplement Dr. Culverhouse's previous training in mathematics and epidemiology with training in statistics and molecular biology. Experience from this period of collaboration, coursework, and mentoring will enable Dr. Culverhouse to play a leading quantitative role in multi-disciplinary scientific teams.