Classification performances comparison. Figure (a) shows the classification results for the breast cancer data. The X-axis represents the percentage of samples that are classified using the expression data and the Y-axis represents the corresponding accuracy at that point. For example, the third point from the left in the black curve denotes, 75% accuracy is achieved by classifying 20% percent of samples using the expression data. Clinical(RF) denotes the accuracy curve from the RF algorithm using the clinical data. Genomic(Plsrf-x) denotes the accuracy curve from the plsrf using the expression data. Step with rank(RF+Plsrf-x) denotes the accuracy curve from our approach. Plsrf-xz denotes the accuracy curve from  and last three denote to the accuracy curves from  with three different feature extraction criteria. Figure (b) shows the results from the CNS cancer data with the same algorithm settings as in (a). Figure (c) shows the result from the breast cancer data when Plsrf with PV is applied to the expression data; the algorithm used for the clinical remains unchanged. Since the result from IntegrativeME is not available for the Plsrf with PV setting, here we only compare our approach with the one from . The last Figure (d) corresponds to the result from the breast cancer data. Here GLM is applied to the clinical data and the algorithm for the expression data remains unchanged.