基于cis-SNP的多位点关联性分析方法比较及其在基因表达中的应用

摘要

Objective The association between common variation and rare variation and gene expression was tested by different SNP set methods,and an efficient and robust statistical method was expected.Methods By comparing Burden,SKAT, SKAT-O,MiST,ReLRT and GGRF methods,a simulation study was conducted by using genotype and gene expression data from GEUVADIS to compare the type I error rate,statistical efficiency and computing time of different methods.Based on the com-mon and rare cis-SNP,the ability to identify expressed genes by various methods was further compared by GEUVADIS data.Re-sults The simulation shows that all kinds of methods can control the type I errors in different number of variable points;the sta-tistical efficiency of ReLRT and MiST is the highest,and the operation time of MiST is the shortest.In GEUVADIS data analy-sis,ReLRT found the most expressed genes.Conclusion All kinds of test methods can effectively control type I errors;simula-tion and case data showed that ReLRT has higher test efficiency,but the calculation speed is slower.%目的 比较不同SNP集合方法检验常见变异和罕见变异与基因表达的关联,期望找到高效和稳健的统计方法.方法 对比Burden、SKAT、SKAT-O、MiST、ReLRT和GGRF等方法,通过GEUVADIS数据基因型和基因表达数据建立模拟研究,比较不同方法的一类错误率、统计效能及运算时间.基于常见和罕见cis-SNP,进一步通过GEUVADIS数据比较各种方法鉴别表达基因的能力.结果 模拟显示各种方法在不同变异位点个数下均能很好控制第一类错误,其中ReLRT和MiST的统计效能最高,MiST运算时间最短.在GEUVADIS数据分析中,ReLRT找到的表达基因最多.结论 各种检验方法基本都能有效控制第一类错误.模拟和实例数据显示ReLRT具有更高的检验效能,但计算速度较慢.