The Study And Appplication Of Genetic Variance Component Model Based On Cox Model

Complex human diseases are determined by an array of genetic and environmental factors. The analysis of familial aggregation is the first step for exploring causes of complex human diseases in the genetic epidemiology. The traits between familial members are not independent. The marginal model or generalized linear mixed effects model can be chiefly used to solve the correlated problem in family for common quantitative traits and qualitative traits. However, such as age at onset, age at menarche and other censored traits could be studied in the basis of Cox proportional hazards model, and the model had been often used in the survival analysis. The familial correlation between familial members in complex traits are composed of additive genetic random effect, dominance genetic random effect and familial shared environmental random effect in classic genetic variance component models, and so assume that the random effects follow a multivariate normal distribution. This study introduced genetic variance component in the basis of Cox model, and constructed Cox genetic variance components with several random effects. It was so difficulty to resolve the intractable multivariate integral for the model due to the hypothesis of multivariate normal distribution for the random effects. The Laplace approximation and penalized likelihood theory were used to resolve the intractable multivariate integral in the study, and programmed the code to implement the variance-covariance matrix of random effects with nuclear families and extended families in the software of R and S-Plus, and accomplished the parameter estimation method of Laplace approximation and penalized partial likelihood in coxme function. And the Cox genetic variance components models were applied on the nuclear families data of Shunde in Guangdong and the extended families data of Taixing in Jiangsu, and the parameter estimation method of Laplace approximation and penalized partial likelihood were used to construct the Cox genetic variance component models of age at onset in primary hepatic carcinoma, in order to illustrate the practical application of the model and parameter estimation method. So the main purpose of this study was to provide a effective and convenient statistical method to identified the familial aggregation of a censored trait for genetic epidemiologist.