Skip to Main Content
Microarray technology is a popular and informative technique widely used in experimental molecular biology, which can produce quantitative expression measurements for thousands of genes in a single cellular mRNA sample. Analysis methods for determining significant genes are essential to extracting information from the multitude of data generated from a single microarray experiment. Raw gene expression measurements alone, most often, do not indicate significant genes for the given condition. While analysis methods are abundant, there is a need for enhanced performance when attempting to identify significant genes from such experiments. Additionally, the ability to more accurately predict informative genes from cross-laboratory and/or cross-experiment data can certainly aid in disease detection. We propose the application of combinatorial fusion analysis (CFA) in order to enhance and expedite the identification of significant genes in a cross-experiment analysis. Previous methods to identify significant genes applied SAM to analyze the data sets and then took the intersection of top ranked genes. In this paper, we used CFA to combine the scoring functions of two data sets produced by SAM. Moreover, both score and rank combinations are used. Both combinations can achieve better results than the previous approach of taking the intersection. In addition, by using the rank-score characteristic function as a diversity measure, we are able to show that rank combination performed better than score combination. CFA can robustly identify significant genes from multiple microarray data sets so that experimental biology researchers can efficiently perform the next phase of analysis on a smaller subset of genes.