STATISTICAL ANALYSIS OF FAMILY DATA IN CANCER RESEARCH

这项拨款申请的目的是制定一个全面的 分析框架，家庭数据分析的回归模型， 并促进这种新的生物统计方法的转移， 癌症遗传流行病学 许多关于癌症遗传流行病学的研究都是以家族为基础的 两代或更多代。这种家庭研究包括评估 考虑到特定生物学因素的衡量风险因素的作用 关系，分布和家庭关系的确定 发病年龄、分离连锁和多效性分析中的相关性 以确定主要基因可能参与的病因， 癌症，以及确定某些模型的适用性 致癌作用。理想的统计框架包括依赖于 二元和生存结果与回归变量，但范围 应包括建模的依赖性没有或参考主要 根据遗传学规律，基因在家族中传递。 分析框架是回归模型， 通过指定人的 表型和一组解释变量，包括他的基因型， 关于特定基因座，年长亲属的表型，以及 环境和生活方式的协变量。这种方法的吸引力在于， 它同时提供了重要的 基因和那些由复杂模式的残余家族性 相关性，包括兄弟姐妹，配偶配偶和父母后代 表型相关性，没有或参考明确的遗传或 环境因果机制。已知的致癌机制是 通过选择合适的参数化。 在具体目标I中，模型的表述被扩展到包括 广泛适用的分布和推理方法。然后将这些数据 在具体目标2中进行了调整，以做到以下几点： 癌症家族聚集性及其危险因素的建模与分析 发病年龄的分布及家族相关性; 分离分析; 连锁和多效性分析。 为便利技术转让（具体目标3）， 将开发软件并举办培训讲习班。

The purpose of this grant application is to develop a comprehensive analytic framework, the regressive models for the analysis of family data, and to facilitate the transfer of this new biostatistical methodology to cancer genetic epidemiology. Many studies on the genetic epidemiology of cancer are based on families of two or more generations. Such family studies include the assessment of the role of measured risk factors taking into account specific biologic relationships, the determination of the distribution and familial correlations in age-of-onset, segregation linkage and pleiotropy analyses to determine the possible involvement of major genes in the etiology of the cancer(s), and the determination of the applicability of some models of carcinogenesis. The desired statistical framework encompasses dependent binary and survival outcomes with regression variables, but the scope should include modeling the dependence without or with reference to major genes transmitted in families according to the laws of genetics. The analytic framework is the regressive models which account for familial correlations by specifying a regression relationship between a person's phenotype and a set of explanatory variables including his genotype with respect to specific loci, the phenotypes of older relatives, and environmental and life-style covariates. The appeal of this method is that it simultaneously provides for the effects resulting from important gene(s) and those resulting from complex patterns of residual familial correlations, including sib-sib, spouse-spouse, and parent-offspring phenotypic correlations, without or with reference to explicit genetic or environmental causal mechanisms. Known mechanisms of carcinogenesis are incorporated by choosing suitable parametrization. In specific aim I, the formulation of the models is extended to include broadly applicable distributions and methods of inference. These are then adapted in specific aim 2 to do the following: Modeling and analysis of familial aggregation of cancer and risk factors; Distribution and familial correlations in age-of-onset; Segregation analysis; Linkage and pleiotropy analysis. To facilitate the technology transfer (specific aim 3), appropriate software will be developed and training workshops will be organized.