Abstract
The growing popularity of microarray technology for testing changes in gene expression has resulted in multiple laboratories independently seeking to identify genes related to the same disease in the same organism. Despite the uniform nature of the technology, chance variation and fundamental differences between laboratories can result in considerable disagreement between the lists of significant candidate genes from each laboratory. By adjusting for known differences between laboratories through the use of covariates and employing a Bayesian framework to effectively account for between-laboratory variability, the results of multiple similar studies can be systematically combined via a meta-analysis. Meta-analyses yield additional information not available from any single study and provide a clearer understanding of each gene’s true relationship to the disease of interest. A simulation model based on the Barley Affymetrix GeneChip microarray demonstrates the utility of this approach. Further illustration is provided from a mouse model for multiple sclerosis.
Keywords
microarray, meta-analysis, hierarchical Bayes linear model
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.
Recommended Citation
Stevens, John R. and Doerge, R. W.
(2005).
"A BAYESIAN AND COVARIATE APPROACH TO COMBINE RESULTS FROM MULTIPLE MICROARRAY STUDIES,"
Conference on Applied Statistics in Agriculture.
https://doi.org/10.4148/2475-7772.1141
A BAYESIAN AND COVARIATE APPROACH TO COMBINE RESULTS FROM MULTIPLE MICROARRAY STUDIES
The growing popularity of microarray technology for testing changes in gene expression has resulted in multiple laboratories independently seeking to identify genes related to the same disease in the same organism. Despite the uniform nature of the technology, chance variation and fundamental differences between laboratories can result in considerable disagreement between the lists of significant candidate genes from each laboratory. By adjusting for known differences between laboratories through the use of covariates and employing a Bayesian framework to effectively account for between-laboratory variability, the results of multiple similar studies can be systematically combined via a meta-analysis. Meta-analyses yield additional information not available from any single study and provide a clearer understanding of each gene’s true relationship to the disease of interest. A simulation model based on the Barley Affymetrix GeneChip microarray demonstrates the utility of this approach. Further illustration is provided from a mouse model for multiple sclerosis.
