贝叶斯因子法在畜禽全基因组关联分析中的应用

Application of Bayesian Factor Method in Genome-wide Association Studies in Animal

为了剔除畜禽全基因组关联分析结果中的假阳性结果,寻找最优的假设检验方法,解决畜禽全基因组关联分析中的多重比较问题,本研究将现有GWAS研究中常用的七种假设检验方法和贝叶斯因子法进行比较。通过对模拟数据和公开数据集的研究,结果表明:畜禽全基因组关联分析中用贝叶斯因子法进行假设推断,其优良的统计性能与假设检验数目(SNP数)和最小等位基因频率(MAF)基本无关,其在假设检验中的某些表现优于其它几种基于p值(p-value)的方法。本研究为进一步解决畜禽全基因组关联分析中的多重比较问题奠定了基础。

In order to eliminate the false positive results in the analysis of genome-wide association of livestock and poultry, we attempted to find the optimal solution of the hypothesis test method to the multiple comparison problem in the studies of genome-wide association of livestock and poultry. Seven kinds of hypothesis test methods commonly used in the study of GWAS and Bayesian factor method were compared in this research. The study of the simulation data and public data set showed that the statistical properties of the hypothesis test and excellent number(SNP number) and the minor allele frequency(MAF) was almost independent based on hypotheses inferred by Bayesian factor method of genome-wide association analysis of livestock and poultry. Some of the performance in Bayesian factor method of genome-wide association analysis should be better than that of others based on the p value(p-value) method. This study might lay a foundation for further solving the multiple comparison problem analysis of genome-wide association in livestock and poultry.