Using biologic knowledge to discover molecular correlations between human renal cell carcinoma pathways

451 Background: Renal cell carcinoma (RCC) is known to be resistant to chemotherapy. There is need for the identification of biomarkers capable to determine RCC prognosis factors and metastatic potential obtainable from non-invasive or minimally invasive techniques. Our aim was to derive predictive models which could predict more accurately than any one factor alone. Methods: To studythe cascade of events leading to the formation and progression of RCC, we assessed 29 markers by immunohistochemistry and qRT-PCR using tissue micro-array (TMA). Results: Multivariate logistic regression showed the best proteins combination for node status (NOTCH1 and GLUT5) and pelvis invasion (EGFR and DLL3). ROC curve analyses were made to analyse the accuracy of the best candidate proteins; it should be noted NOTCH1 and GLUT5 for node status prediction (AUC=0.833, 95% CI, 0.744-0.922; p