Statistical Methods for Gene Mapping Based on a Confidence Set Approach

在基因研究的许多领域，多重测试是一个重要但困难的统计问题。当在整个基因组中筛选许多标记以确定它们与疾病位点的联系或关联时，就会出现一个特殊的多重测试问题，这是本项目的重点，其广泛的长期目标是开发适用于遗传和基因组研究各个领域的方法。拟议方法的主要主旨在于它提出了关于联系的假设。传统的链接假设通常是零假设为无链接，替代假设为链接。在新的提法中，零假设和替代假设正在“颠倒”，零假设是紧密联系，替代假设是松散联系或不联系。这种新模式的两个基本优势是：第一，在基因组扫描研究中进行的测试的数量不需要进行多样性调整，第二，即使在初步的基因组扫描研究阶段，疾病基因的位置也可以缩小到一个小的基因组区域。第一个具体目标是开发基于假设的参数检验来构建疾病基因位置的标记或可信区域(区间)的置信度集的方法。将开发单标记和多标记方法来处理来自一般家系的数据。第二个具体目标可以被视为第一个目标的非参数对应。将开发基于使用等位基因共享统计的非参数检验来构建置信集/区域的方法。将考虑各种各样的等位基因共享统计和数据类型，从简单的结构(同胞、亲属对)到一般的家系。随着人类基因组计划的完成，以及高通量基因分型技术的发展，现在搜索分布在整个基因组中的多达数千个遗传标记来寻找疾病易感基因已经成为一件例行公事。该项目旨在开发适合于探测这些标记中的每一个的统计方法，而不影响找到附近易感基因的能力。随着参与的家系数量的增加，错误地暗示一个标记不在疾病基因较短距离内的比率最终将降至零。这不仅增加了成功识别疾病基因的机会，而且还减少了追逐“幽灵”基因的机会，从而节省了大量资源。因此，开发的方法可以成为基因作图界的一个有价值的工具。特别是，预计本项目中开发的方法将应用于研究人员正在与医生和其他研究人员合作的项目中关于一系列自身免疫性疾病的数据，包括系统性红斑狼疮和多发性硬化症。

Multiple testing is an important but difficult statistical issue in many areas of genetic research. One particular multiple testing problem arises when many markers are screened throughout the genome for their linkage or association with a disease locus, which is the focus of this project, with a broad long-term objective of developing methods applicable in various areas of genetic and genomic research. The main thrust of the proposed approach lies in its formulation of the hypotheses for linkage. Traditionally, hypotheses for linkage are usually set up with the null hypotheses being no linkage and the alternative hypothesis being linkage. In the new formulation, the null and alternative hypotheses are being ``reversed'', with the null hypothesis being tight linkage and the alternative hypothesis being loose linkage or no linkage. Two of the fundamental advantages with this new paradigm are: first, multiplicity adjustment is unnecessary for the number of tests performed in a genome-scan study, and second, the location of a disease gene can be narrowed down to a small genomic region, even at the stage of a preliminary genome-scan study. The first specific aim is to develop methods for constructing confidence sets of markers or confidence regions (intervals) of disease gene locations based on parametric tests of the hypotheses. Single-marker and multiple-marker approaches will be developed for data from general pedigrees. The second specific aim can be viewed as a non-parametric counterpart of the first aim. Methods will be developed for constructing confidence sets/regions based on non-parametric tests using allele-sharing statistics. A wide variety of allele-sharing statistics and data types, ranging from simple structures (sibships, relative pairs) to general pedigrees, will be considered.With the completion of the Human Genome Project, and the development of high throughput technology for genotyping, it is now a routine matter to search up to thousands of genetic markers distributed throughout the genome to look for disease susceptibility genes. This project aims at developing statistical methods suitable for probing each of these markers without compromising the power of finding nearby susceptibility locus. As the number of participating families increases, the rate of falsely implicating a marker not located within a short distance from a disease gene will eventually go down to zero. This would not only increase the chance of successful identification of disease genes, but would also save tremendous resources by reducing the chance of going after "ghost" genes. Thus, the methods developed can be a valuable tool to the gene mapping community. In particular, it is expected that the methods developed in this project will be applied to data from projects, on which the investigator is collaborating with medical doctors and other researchers, on a range of autoimmune diseases, including systemic lupus erythematosus and multiple sclerosis.