Exploration of empirical Bayes hierarchical modeling for the analysis of genome-wide association study data
Exploration of empirical Bayes hierarchical modeling for the analysis of genome-wide association study data
复制标题
DOI：
10.1093/biostatistics/kxq072
复制
发表时间：
2011-07-01
期刊：
影响因子：
2.1
通讯作者：
Gill, Michael
中科院分区：
文献类型：
--
作者：
Heron, Elizabeth A.;O'Dushlaine, Colm;Gill, Michael
在全基因组关联（GWA）数据的分析中，目的是检测单核苷酸多态性（SNP）与感兴趣的疾病或性状之间的统计学关联。这些SNPs，或它们所涉及的基因组的特定区域，然后被考虑用于进一步的研究。我们通过一个全面的模拟研究表明，包括额外的，生物相关的信息，通过2级经验贝叶斯分层模型框架提供了一个更强大的方法检测相关的SNP。经验贝叶斯方法是一种分析数据的客观方法，无需设置主观参数估计。该框架通过减少通常效应估计值的变异性，提供了更稳定的效应估计值。我们还展示了包括额外的信息，是不是信息和检查功率和假阳性率的后果。我们将该方法应用于一些包含额外生物信息的全基因组关联（GWA）数据集。我们的研究结果与以前的研究结果一致，并在一个数据集（克罗恩病）的情况下，建议一个额外的区域的利益。
In the analysis of genome-wide association (GWA) data, the aim is to detect statistical associations between single nucleotide polymorphisms (SNPs) and the disease or trait of interest. These SNPs, or the particular regions of the genome they implicate, are then considered for further study. We demonstrate through a comprehensive simulation study that the inclusion of additional, biologically relevant information through a 2-level empirical Bayes hierachical model framework offers a more robust method of detecting associated SNPs. The empirical Bayes approach is an objective means of analyzing the data without the need for the setting of subjective parameter estimates. This framework gives more stable estimates of effects through a reduction of the variability in the usual effect estimates. We also demonstrate the consequences of including additional information that is not informative and examine power and false-positive rates. We apply the methodology to a number of genome-wide association (GWA) data sets with the inclusion of additional biological information. Our results agree with previous findings and in the case of one data set (Crohn's disease) suggest an additional region of interest.