A plasma proteomics-based model for predicting response to neoadjuvant chemotherapy in ovarian cancer.

Abstract

3056 Background: Neoadjuvant chemotherapy (NACT) is a standard treatment option for advanced high-grade serious ovarian cancer (HGSOC). Following interval debulking surgery, pathologists assess tumor response to initial chemotherapy using a standardized chemotherapy response score (CRS) corresponding with patient survival. We sought to determine the feasibility of developing a proteomic-based biomarker to predict response to NACT based on pre-treatment plasma samples. Methods: Pre-treatment samples were collected from 71 HGSOC patients receiving platinum-taxane combination NACT. Deep plasma proteomic profiling was performed using SomaLogic’s 7K aptamer-based technology. Based on the proteomic profiles, a computational model was developed to predict CRS, focusing on differentiating between poor CRS (CRS1) versus partial or near-complete CRS (CRS2/3). The model counted the number of response-associated proteins per patient. Patient scores were derived by resampling into training and test sets, averaging test set results. Patients were stratified into groups (i.e., 'Chemo-responsive' or 'Chemo-resistant') based on median score. Bioinformatic analysis of the HGSOC-specific proteomic biomarkers was performed to gain insight into the potential mechanisms driving NACT therapeutic benefit and resistance. Results: Our proteomics-based predictive model differentiated between patients with CRS1 versus CRS2/3 (ROC AUC = 0.67, p = 0.008). CRS association with disease-free survival (DFS) and overall survival (OS) in this cohort was consistent with a previous meta-analysis, though not reaching statistical significance (CRS3 versus CRS1/2, HR = 0.62, 95% CI: 0.27-1.42, p = 0.25 for DFS; HR = 0.67, 95% CI: 0.21-2.09, p = 0.48 for OS). Hazard ratios for patient classification as ‘Chemo-responsive’ versus ‘Chemo-resistant’ trended in the same direction (HR = 0.70, 95% CI: 0.37-1.30, p = 0.25 for DFS; HR = 0.65, 95% CI: 0.25-1.72, p =0.39 for OS). The predictive model incorporated 62 proteins. Of these, 27 were elevated in the plasma of patients with poor CRS compared to those with partial or near-complete CRS. These proteins were notably enriched in pathways related to cell death resistance. Moreover, several of these elevated proteins have previously been linked to HGSOC, particularly in the context of chemotherapy resistance. Conversely, patients with partial or near-complete CRS showed significant enrichment of proteins associated with genome instability and mutation. Conclusions: This study demonstrates the feasibility of CRS prediction using plasma proteomics at baseline, potentially complementing existing imaging approaches. While current treatment options limit immediate clinical utility, these findings provide novel insights into biological determinants of chemotherapy response and may become predictive biomarkers for novel treatment protocols.