Marker Gene Selection Evaluation in Brain Tumor Patients Using Parallel Coordinates

Based on microarray gene expression datasets, many statistical methods have been proposed to locate the significant differentially expressed genes (marker genes) among different sample groups. Although robust models for identifying marker genes more accurately is an area of intense research, effective tools for the evaluation of results is often ignored in the literature. In this paper, we propose a novel visualization method to evaluate the marker gene selection process in brain tumors. We use parallel coordinates to visualize the expression patterns of the marker genes in a way that facilitates the qualitative measure of the overall accuracy of the selection process. We exploit our proposed method to evaluate the robustness of 2 statistical tests as examples of gene selection methods. To measure the reliability of our evaluation process we exploit a brain tumor prediction mechanism based on the selected marker genes. It is anticipated that if the marker genes are precisely located, tumor prediction accuracy is improved. We check if our parallel coordinate based visual understanding about the exactness of the selected marker genes is in agreement with tumor prediction method. We surmise that the prediction results support our conclusion based on visualization.