结合偏最小二乘回归的复杂疾病基因定位全基因组关联分析

发布时间：2018-03-24 13:31

本文选题：偏最小二乘回归　切入点：多元统计分析　出处：《浙江大学》2010年硕士论文

【摘要】： 人类常见复杂遗传疾病的基因定位数年来一直是生物学和遗传学研究的热点领域之一。随着DNA测序技术和芯片技术的不断发展,多元统计分析,数据挖掘和机器学习等多种分析方法被广泛应用于全基因组关联分析的研究中,而全基因组关联分析已经成为搜索与疾病易感性相关遗传变异的主要方法。但是,大多数分析方法都基于不同的遗传变异之间功能相互独立的假设,而遗传变异对于疾病易感性的影响主要还是依靠基因-基因与基因-环境的交互作用。忽略遗传背景可能会降低检测效能与研究的可重复性,为了克服基于相互独立性假设的全基因组关联分析中的缺陷,我们使用多元统计分析中的偏最小二乘回归方法的全基因组关联分析对人类常见复杂疾病数据进行研究。我们将偏最小二乘回归应用于WTCCC的七种人类常见复杂疾病的真实数据,并检测出一个新的类风湿关节炎的相关SNP位点,一个新的2型糖尿病相关SNP位点以及一个在以往的研究中只有中等显著性但是在偏最小二乘回归研究中有较高显著性的高血压相关SNP位点rs2820037。我们没有找到在以往的研究中有较高显著性的双向障碍相关SNP位点rs420259,这与近期的研究中报道此SNP位点是无法重复的结果相吻合。此外,我们将偏最小二乘回归方法与传统的单SNP位点统计分析,逐步Logistic回归分析和主成分回归分析进行比较,发现前者具有较为明显的效能优势。
[Abstract]:Gene mapping of common complex genetic diseases in humans has been one of the hot research fields in biology and genetics for several years. With the development of DNA sequencing technology and chip technology, multivariate statistical analysis, Many analysis methods, such as data mining and machine learning, have been widely used in the study of genome-wide association analysis, and genome-wide association analysis has become the main method to search for genetic variation associated with disease susceptibility. Most analytical methods are based on the assumption that the functions of different genetic variations are independent of each other. The effect of genetic variation on disease susceptibility mainly depends on the interaction between gene and gene environment. Neglecting genetic background may reduce the detection efficiency and the repeatability of research. To overcome shortcomings in genome-wide association analysis based on the assumption of mutual independence, We use partial least square regression method in multivariate statistical analysis to study human common complex disease data. We apply partial least squares regression to the real data of seven common human complex diseases in WTCCC. A new SNP locus associated with rheumatoid arthritis was detected. A new type 2 diabetes associated SNP locus and a moderately significant SNP locus rs2820037, which were only moderately significant in previous studies, were found in the partial least squares regression study. The SNP locus rs420259 is highly significant in this study, which is consistent with the results reported in recent studies that the SNP locus is unrepeatable. We compared the partial least square regression method with the traditional statistical analysis of single SNP locus, stepwise Logistic regression analysis and principal component regression analysis, and found that the former has obvious efficiency advantage.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2010
【分类号】：R346

【相似文献】