Statistical tools for phenotype definition in genetic studies

点击翻译按钮获取中文摘要

Phenotype definition is a necessary prerequisite to establish reliable genotype-phenotype relationships in genetic studies. In order to identify candidate regions of interest for complex diseases, a "well defined" phenotype should have the following characteristics: i) be highly heritable; ii) be genetically homogeneous; iii) combine the information collected on a variety of items related to the disease. The goal of this proposal is to establish a set of statistical tools for defining the phenotype in genetic studies, when a large number of items are collected on the patients and/or when a genetically relevant disease definition is unclear, as is the case in psychiatric disorders for example. Based on previous work, I will try to focus on the three following objectives: 1) Selection of the phenotype based on its heritability (in family studies). I am planning to extend the principal component approach based on heritability for combining phenotype information (Ott, 1999). This approach will be extended to account for any type of family structure. 2) Construction of phenotypes based on a factorial analysis of the items collected (in family studies). This project is a direct extension of the latent class model we developed for familial/pedigree data (Labbe et al., 2009; Tayeb et al., 2009). The proposed project consists of extending this work to the continuous cases, i.e. where the disease is controlled by unobserved latent continuous factors instead of categorical classes. 3) Association study for secondary phenotypes (case-control studies). This project will focus on the statistical approach used to analyze secondary phenotypes, i.e. phenotypes (traits) that have been collected in addition to the disease trait in case-control studies. Because of unequal selection probabilities between cases and controls, these secondary phenotypes are not collected from a random sample of the population, and standard methods analyzing this type of phenotypes can lead to misleading results (Lin & Zeng, 2009). I am planning to propose a joint model for the disease status and the secondary phenotype, which will account for the association between the disease and secondary trait, and between the secondary trait and a given SNP.