Application Of Decision Tree Model In Whole Genome Association Of Coronary Heart Disease

Coronary artery heart disease (CAHD) is a complex disease that is caused by both environmental and genetic factors and has become one of the leading causes of death and disability worldwide. With the development of the research of genome-wide association study (GWAS), the research of the cause of CAHD is emphasizing particularly on gene and becomes hotspot recently.Plenty of statistical methods are used in GWAS, such as Chi-square test, logistic regression, analysis of variance and so on. Some multiple testing corrections include Bonferroni correction, step-down Bonferroni correction, permutation correction and etc. All these are exploited to two kinds of problems. First is to find the genetic susceptibility loci and second is to forecast whether to be attacked by CAHD or not. However, due to the huge information of gene, the classical statistical parameter models will bring bias. Furthermore, the methods they used cannot solve the two kinds of problems at the same time.The thesis aims at the data from WTCCC, which contains4864observations (1926CAHD) and393473genetic loci and use the decision tree to pick out5genetic susceptibility loci (independent variables). Meanwhile we can predict whether to be attacked by CAHD or not by using the decision tree model. The classification correct rate of both training sample and validation sample is about74%stably. Additionally, the leaf nodes of the decision tree reveal the proportion of the CAHD sufferers to the all observed in the parent node, which also represents the prediction of the disease hazard and gives a good screening of the CAHD high-risk groups.