基于模型的多因子降维方法在基因-基因/环境交互作用分析中的应用

Detecting gene-gene/environment interactions by model-based multifactor dimensionality reduction

介绍一种基于模型的多因子降维方法（MB-MDR），并通过实例说明其分析流程及其在基因一基因／环境交互作用分析中的应用。结果显示该方法可用于原始样本量较小的资料研究，同时也能解决许多经典MDR方法的不足；与其他MDR扩展方法相比，在探索交互作用方面具有更高的统计效能，并已成功应用于膀胱癌、湿疹等研究。MB-MDR能够处理二元性状和数量性状，并可在模型中调整因子的边际效应和混杂因子，与其他非参数方法相比具有一定优势。因此MB-MDR在基因一基因／环境交互作用分析中具有较好的应用前景。

This paper introduces a method called model-based multifactor dimensionality reduction （MB-MDR）, which was firstly proposed by Calle et al., and can be applied for detecting gene-gene or gene-environment interactions in genetic studies. The basic principle and characteristics of MB-MDR as well as the operation in R program are briefly summarized. Besides, the detailed procedure of MB-MDR is illustrated by using example. Compared with classical MDR, MB-MDR has similar principle, which merges multi-locus genotypes into a one-dimensional construct and can be used in the study with small sample size. However, there is some difference between MB-MDR and classical MDR. First, it has higher statistical power than MDR and other MDR in the presence of different noises due to the different way the genotype cells merged. Second, compared with MDR, it can deal with all binary and quantitative traits, adjust marginal effects of factors and confounders. MB-MDR could be a useful method in the analyses of gene-gene/environment interactions.