An across Factor Normalization Based SVD Approach to Analysis of Gene Expression Profiles for Uncovering Biomarkers in Ovarian Carcinoma Chemotherapy Responses

Analyzing microarray data to identify interesting genes is a well-established methodology but often results in inconsistent conclusions and even fails because of the variations of experimental conditions. This study proposes an across factor normalization based singular value decomposition approach to microarray data analysis. The approach has been applied to analyze gene expression profiles to identify differentially expressed genes associated with Ovarian Carcinoma chemotherapy responses. The influences of experimental conditions are identified by correlations analysis and Friedman'M test and illustrated by Scatter Plots. These influences are then removed by across factor normalization. Experimental results showed that after across factor normalization, the samples associated with different chemotherapy responses are significantly clustered together and genes linked to differential chemotherapy responses are identified.