探讨χ~2检验结合FDR筛选致病SNPs位点的适用条件

Exploration of the Applicable Conditions of χ~2 Test Combined with False Discovery Rate for Screening Pathogenic SNPs

目的在单核苷酸多态性(SNPs)数据中探讨不同模拟条件χ2检验结合错误发现率(FDR)筛选SNPs位点的适用条件。方法依据2009年2月发布HapMapⅢ期美国犹他州北欧和西欧后裔人群22号染色体前5 000个SNPs数据,采用HAPGEN2模拟病例对照数据,运用Haploview4.2筛选标签SNPs(TagSNPs),比较不同模拟条件筛选致病SNPs的正确率。结果相对危险度(RR)获取方式无显著差异;3种遗传模型均表现正确率随RR值增大而增高,RR相同时,加性模型正确率最高,显性模型次之,隐性模型最低;加性模型RR>2.2、显性模型RR>4和隐性模型RR>5时,正确率超过60%。结论χ2检验结合FDR在加性模型效果最佳,实际科研工作需依据目标疾病具体情况考虑是否适合χ2检验结合FDR方法。

Objective To explore the applicable conditions of X^2 test combined with false diseovery rate （FDR） for screening single nucleotide polymorphisms （SNPs） locus by investigating the different simulation conditions in the SNPs data. Methods According to the data o{ the first 5,000 SNPs in chromosome 22 of Utah residents with Northern and Western Euro- pean ancestry from the CEPH collection released in HapMap !II in February 2009, case - control data were simulated by HAP- GEN2, and Tag SNPs were screened by Haploview4.2. The correct rates of different simulation conditions in screening patho- genic SNPs were compared. Results The methods of obtaining the relative risk （RR） were not significantly different. Three genetic models all showed that the correct rate increased with the rising RR. For the same RR, the additive model had the high- est correct rate, the dominant model had the second, and the recessive model had the lowest. For the additive model with RR 〉 2.2, the dominant model with RR 〉4 and the recessive model with RR 〉 5, the correct rate was over 60 %. Conclusions Combination of X^2 test with FDR has the best application in the additive model. In the actual scientific research, the feasibility of X^2 test combined with FDR should be considered on the basis of the concrete conditions of the target disease.