The evolving understanding of the human genome sequence and recent advances in genomic technologies provide a strong foundation to resolve the genetic basis of common human diseases. However, common diseases that cluster in families, such as many cancers, cardiovascular, and diabetes, have much heterogeneity in their etiology and their phenotypic expression. Sources of this heterogeneity range from genetic, to phenotype, to environmental and behavioral risk factors, to study design, to a combination of all of these sources. The success of linkage studies for common diseases has been limited, and future success may hinge on the ability to account for (and control) a wide range of heterogeneity. The goals of this proposed research are to improve the ability to identify and characterize susceptibility genes for complex human traits by developing statistical and computational methods that account for linkage heterogeneity.
Aim 1. Recursive Partitioning Trees: To improve identification of susceptibility genes that play critical roles in subsets of families, we will develop new quantitative methods for evaluation of genetic linkage heterogeneity across subsets based on recursive partitioning trees. These new developments, based on modem statistical and computing algorithms, should improve the power to detect linkage in the presence of a large amount of heterogeneity caused by non-genetic factors that influence disease in non-linear ways.
Aim 2. Regression Models: To increase the power to detect linkage for complex genetic traits, and refine the regions of promising linkage, we will develop new statistical methods that provide flexible ways to directly model the influence of critical factors on identity-by-descent sharing probabilities for genetic linkage studies.
Aim 3. User-friendly software: User-friendly software that implements the proposed methods, including well-documented procedures and examples of their usage, will be provided free to the scientific community.
