CLASSIFICATION METHODS FOR DETECTING DISEASE LOCI

Washington University, Saint Louis MO

Investigators

Linked publications & trials

Abstract

DESCRIPTION (Adapted from the Investigator's Abstract): Bold steps must be taken to advance our understanding of the genetic and associated co- variates affecting the inheritance of complex diseases. To that end, this proposal will develop improved quantitative methods to detect genetic factors contributing to increased susceptibility to complex disorders and implement these methods in software for distribution to the research community. The methods will concentrate on the use of classification techniques applied to allele sharing data and other risk factors which affect the trait. Allele sharing methods for mapping genes will be extended to include the classification methods known as latent class models, cluster analysis, and artificial neural networks, as well as a novel use of logistic regression Co-variates such as gender, parental diagnosis, or other concomitant factors will be systematically studied through applications to both stimulated and existing data sets. An additional goal is to determine the optimal distribution of relative pairs (e.g. siblings, first cousins) for these methods. Of great importance to this proposal is the development of well-documented, user-friendly software and documentation which will be distributed to the scientific community via the Internet. Existing software developed by the PI will be extensively expanded for latent class models. Existing cluster analysis software will be modified and combined for ease of use. This proposal consists of theoretical exploration, computer simulation, data analysis, and software development. First, solutions of theoretical questions relating to classification techniques will be pursued; second, adaptation of computer programs to implement the analytic methods, and investigation into alternative research strategies will be accomplished. The new strategies will be applied to stimulated data, and finally, to existing data sets of pedigrees in which a complex trait has been diagnosed. Findings from this research may contribute to the ability to locate susceptibility loci in complex traits and to the clarification of those etiological mechanisms responsible for susceptibility.