Dimension reduction and mixed-effects model for microarray meta-analysis of cancer

The rapid advances in high-throughput microarray technologies greatly facilitate the disease biomarker discovery. However, the potential of these microarray data has not yet been fully utilized. This is partly due to the limited sample sizes of each individual study. Combining microarray data from multiple studies improves the statistical power of detecting differentially expressed genes. Here we present a method for combining the microarray datasets at array probeset level. Using datasets from two commonly used array platforms, the Affymetrix Human Genome U133A and Human Genome U133 Plus 2.0 arrays, we found laboratory effects may be more influential than the platform effect. A visualization scheme for merging the array data from different array platforms was proposed to qualitatively judge the degree of agreement between datasets. A mixed-effects model was applied to identify differentially expressed genes from the merged array data.