An accurate estimate of disease risk is critical to clinical management of individuals with a missense mutation of a disease susceptibility gene. In practice, it is a challenge to accurately assess a missense mutation's contribution to disease susceptibility as sufficient sample size or an appropriate functional assay are often not available. For cancer susceptibility genes, measures of disease association will in the foreseeable future be based on pedigree data collected in clinics specializing in individuals at high risk of genetic susceptibility to cancer. Absent a correction, this sampling mechanism is likely to lead to an overstatement of the mutation's contribution to disease. In this study, we propose a Bayesian hierarchical model to study the disease causality of missense mutations given pedigree data collected in the high risk setting. The model classifies missense mutations as deleterious or non-deleterious. The hierarchical structure of the model makes the systematic comparison of the effects of different missense mutations possible, allowing us to study them as a group instead of one at a time. In addition to the missense pedigrees, the model utilizes a group of pedigrees identified through probands tested positive for known deleterious mutations and a group of test-negative pedigrees, both obtained from the same clinic, to calibrate the classification and control for potential ascertainment bias. We apply this model to study missense mutations of breast--ovarian susceptibility genes BRCA1 and BRCA2, using data collected at the Duke University Medical Center in Durham, North Carolina.

Keywords: Disease Gene Characterization, Bayesian Classification, Ascertainment Correction.

The manuscript is available in PostScript and PDF formats.