利用精神分裂症資料來比較五種常用的偵測基因基因交互作用效果的方法

Abstract

有越來越多的證據顯示，基因基因交互作用是普遍存在於常見複雜性疾病之中的。為了發現這些交互作用與疾病的相關性，已經發展出了許多的統計方法。我們有興趣的是如何使用這些方法來實際分析資料，並且想要比較這些方法。在這個研究中，我們利用五種常用的方法：卡方檢定、邏輯迴歸模型（LRM）、 bayesian epistasis association mapping (BEAM) algorithm、classification and regression trees (CART) 以及 the multifactor dimensionality reduction (MDR) method來分析一組精神分裂症的病例－對照研究資料。我們的分析顯示，有一些顯著的單一marker效果以及基因基因交互作用效果是與精神分裂症有高度相關的。在研究的最後部份，我們希望能比較這五種方法在預測疾病狀態上的能力，我們利用cross-validation來比較這五種方法的預測能力。

There are more evidences that gene-gene interaction is probably ubiquitous in complex disease. Several statistical methods have been developed to detecting such association. We are interesting in how to use these methods in a real data and want to compare these methods. In the present study, we applied five commonly used methods: chi-square test, logistic regression model (LRM), bayesian epistasis association mapping (BEAM) algorithm, classification and regression trees (CART), and the multifactor dimensionality reduction (MDR) method to a schizophrenia case-control dataset. Our study show evidence for several single marker effects and gene-gene interactions associated with schizophrenia. At the final part, in order to assess the ability of prediction with these five methods, cross-validation is also proposed along with these methods.