Novel functional proteomics approach to defining ovarian cancer molecular heterogeneity

5502 Background: A number of clinicopathologic risk factors are used for survival prediction and clinical decision-making in epithelial ovarian cancer (EOC). Information from novel technologies such as gene arrays has not had an impact on patient management. We studied EOC protein signaling profiles to determine if their addition to accepted clinicopathologic factors improves their accuracy in predicting individual patient outcomes. Methods: We applied a novel functional proteomics technology, reverse phase protein array (RPPA), to quantify expression and activation of 42 steroid and kinase signaling pathway proteins in 106 high-grade EOCs from patients with stages 1–4 tumors managed with surgery and platinum-based chemotherapy. Cox regression analysis and a novel committee modeling approach were used to study the impact of functional proteomics on patient outcomes. Results: In a Cox model using only clinical variables, stage and residual disease were significantly related to overall survival. By adding the proteins to the clinical Cox model, two proteins that were significantly associated with overall survival on univariate analysis (phosphorylated-MAPK (p-MAPK; log rank p = 0.0047) and progesterone receptor (PR; log rank p = 0.027)) remained significant at the alpha=0.10 level (z-test p-values 0.074 and 0.034, respectively, when treated as binary variables according to martingale residual plots); as a result, these two proteins added to the predictive accuracy of the clinical survival model. However, using the novel committee modeling approach in test and validation EOC sets, a closest neighbor metric was applied to successfully define distinct proteins groups, each composed of nine proteins, that are predictive of specific survival times in patients with EOC. This granular approach to modeling is particularly suited to defining the molecular heterogeneity of EOC. Conclusions: EOC is a complex process with significant individual variability. Using novel approaches to functional proteomic study and statistical modeling, our striking finding is that distinct combinations of steroid and kinase signaling proteins are predictive markers of specific survival times in EOC. No significant financial relationships to disclose.