病例-对照设计的似然比罕见变异关联性检验的构建和模拟评价  

Development and simulations of likelihood ratio test for rare variants association analysis in case control studies

作　　者：曾平[1] 赵杨[2] 陈峰[2] 金英良[1] 张耀东[1] 

机构地区：[1]徐州医科大学公共卫生学院流行病与卫生统计学教研室,江苏徐州221004 [2]南京医科大学公共卫生学院生物统计学系,南京210009

出　　处：《郑州大学学报（医学版）》2017年第2期122-126,共5页Journal of Zhengzhou University(Medical Sciences)

基　　金：国家自然科学基金项目81402765;国家统计局全国统计科学研究项目2014LY112

摘　　要：目的:发展适用于病例-对照设计的似然比罕见变异关联性分析方法。方法:在logistic混合模型的框架下基于PQL算法建立伪数据,将二分类表型转化为连续型表型的关联性分析,然后借助线性混合模型的方差成分检验执行关联性分析。采用Monte Carlo模拟评价该方法的有效性,并与现有方法进行对比。结果:模拟显示,在不同情况下包括似然比检验在内的所有统计检验都能有效控制Ⅰ型错误;在效应方向相同情况下,Burden检验、SKAT-O和Mi ST的统计效能高;在效应方向不同的情况下似然比检验优于其他方法。结论:基于logistic混合模型和PQL算法的似然比检验可有效用于病例-对照设计的罕见变异关联性分析。Aim： To develop a likelihood ratio test for rare variant association analysis in case control studies. Methods： The likelihood ratio test was constructed under the framework of logistic mixed models; a new pseudo-data set was obtained via the working response in the PQL algorithm,which transformed the problem of testing rare variants association in case control studies into the problem of testing rare variants association with continuous under the framework of linear mixed models. The Monte Carlo simulation was conducted to evaluate the proposed test and to compare with existing methods. Results： The simulations showed that all the methods could control the type Ⅰ error correctly. When the effects of the causal rare variants were in the same direction,the Burden test,SKAT-O and Mi ST were most powerful; while if both positive and negative effects were present,the likelihood ratio test outperformed the others. Conclusion： The likelihood ratio test developed using the logistic mixed models and the PQL algorithm can be applicable to rare variant association analysis in case control studies.

关 键 词：罕见变异 关联性分析 logistic混合模型 似然比检验 

分 类 号：R195.1[医药卫生—卫生统计学]