Extraction of Informative Genes from Integrated Microarray Data

We have recently proposed a rank-based approach as a new microarray data integration method. The rank-based approach, which converts the expression value of each sample into a rank value within the sample, enables us to directly integrate samples generated by different laboratories and microarray technologies. In this study, we show that a non-parametric scoring method can be efficiently employed for the rankbased data, and informative genes can be effectively extracted from the integrated rank-based data. To verify the statistical significance of the scoring results from the rank-based data, we compared the distribution of the score statistics to a set of distributions obtained from the randomly column-permuted data. We also validate our methods with experimental study using publicly available prostate microarray data. We compared the informative genes extracted from each individual data to the informative genes extracted from the integrated data. The results show that we can extract important prostate marker genes by directly integrating inter-study microarray data, which are missed in either single analysis.