Post-treatment pathway profiling and platinum resistance in high-grade serous ovarian cancer: Discovery and external validation study.

Abstract

e17602 Background: High-grade serous ovarian cancer (HGSOC) shows initial platinum sensitivity in >70% of patients, but most resist within 18 months. Current baseline biomarkers (CA-125, BRCA) provide limited predictive power. We hypothesized that residual pathway activation in post-treatment samples predicts resistance better than pre-treatment measurements, as chemotherapy selects resistant clones. Methods: Discovery: We analyzed paired treatment-naïve and post-neoadjuvant (NACT) scRNA-seq data from 11 patients (GSE165897), scoring DDR, PI3K, and VEGF pathways against platinum-free interval (PFI). Validation: We analyzed copy number (CN) signatures from 47 paired diagnosis-relapse samples (BriTROC-1), testing if relapse features outperformed diagnosis features. Statistics included Spearman, ROC with bootstrap 95% CIs, and Wilcoxon tests. Results: Discovery: Post-treatment DDR score showed strong inverse correlation with PFI (ρ=-0.711, p=0.014); pre-treatment did not (ρ=-0.182, p=0.592). Post-treatment PI3K achieved AUC=0.750 (p=0.020) for resistance classification. Pathway deltas showed no correlation, indicating absolute post-treatment state predicts resistance. High post-treatment DDR patients had significantly shorter PFI (median 65 vs 393 days, p=0.012). Validation: CN signature 7 at relapse achieved AUROC=0.874 (95% CI: 0.750-0.974, p=0.035), outperforming diagnosis (AUROC=0.694). This +0.180 improvement confirmed the serial profiling hypothesis across modalities. Conclusions: Post-treatment features predict resistance across independent cohorts and modalities. Key insight: Absolute post-treatment molecular state—not baseline or change—is the predictive signal. Strongest signals (DDR, CN signature 7) align with known mechanisms (DNA repair, CCNE1). This approach could enable prediction months before radiographic progression, complementing CA-125 kinetics. Larger prospective studies are needed despite consistent findings.