Abstract A033: Plasma cell-free DNA methylation-based prognosis in metastatic castrate-resistant prostate cancer

Abstract Background: Prognostication in metastatic castration-resistant prostate cancer (mCRPC) is based largely on non-specific proteins and clinical factors which cannot be targeted therapeutically for delaying cancer progression in lethal mCRPC. DNA methylation, which can capture tumor-specific epigenetic changes, shows promise as a platform for identifying biology-based prognostic markers. This study investigates plasma cell-free DNA (cfDNA) methylation to identify mCRPC prognostic markers and develops an integrated clinical-molecular nomogram model for risk stratification and potentially personalized treatment strategies. Methods: Plasma was prospectively collected from localized prostate cancer (PC) (n=19), metastatic hormone-sensitive PC (mHSPC) (n=28), and metastatic castration resistant PC (mCRPC) (n=48) patients (pts). Extracted cfDNA underwent enzymatic methylation sequencing (EM-seq) targeting 366 genomic regions implicated in PC biology. Whole genome sequencing libraries were captured with a Twist targeted methylome panel. Bismark was used for methylation calling. mHapSuite was used to identify methylation haplotype blocks (MHBs) and subsequently calculate methylation haplotype load (MHL), the successive methylation at each fragment length. Differentially methylated regions (DMRs) unique to mCRPC were identified with Welch’s t-tests by comparing PC to mHSPC and mCRPC states. mCRPC survival analysis was performed with Cox proportional hazards models, Kaplan-Meier analysis, and nomograms. Results: From 366 target genomic regions, we identified 316 MHBs (linkage disequilibrium r2>0.5, CpG sites≥5) across the 96 PC patients. Comparing PC to mHSPC, 29 DMRs were identified (p-value<0.05). Comparing mHPSC to mCRPC, 271 DMRs were identified, of which 28 were shared with the localized PC vs. mHSPC analysis (p-value<0.05). A MHB-based composite risk score of the top 22 MHBs was generated for DMRs restricted to mCRPC state which showed worse mCRPC survival in high-risk patients (median survival time=15.5 months) than in low-risk patients (median survival time=36.5 months) (log-rank p-value=0.00052). Combining the top significant clinical markers with the MHB-based composite risk score also demonstrated decreased survival probability in high-risk patients (median survival time=15.5 months) as compared to low-risk patients (median survival time=42.8 months) (log-rank p-value=0.00025). Incorporation of the MHB-based score with predicted ctDNA fraction and clinical biomarkers into a multi-modal nomogram model improved the prognostic performance (6-month AUC=0.99, 1-year AUC=0.90, 2-year AUC=0.87) over clinical biomarkers alone (6-month AUC=0.98, 1-year AUC=0.87, 2-year AUC=0.84). Conclusions: Our findings demonstrate that cfDNA methylation signatures can complement existing prognostic models, offering a tumor-biology based personalized approach to mCRPC prognostication, which also identify therapeutic targets in pts with short survival. These findings warrant further validation in a newly recruited patient cohort. Citation Format: Manish Kohli, Jodie Wong, Yijun Tian, ManishKumar S. Patel, Kapil Avasthi, Claire Hanson, Enos Ampaw, Rebekah Gutowski, Muhammad Zaki H. Fadlullah, Joseph Finklestein, Aik C. Tan, Jong Park, Brandon J. Manley, Chiang-Ching Huang, Liang Wang. Plasma cell-free DNA methylation-based prognosis in metastatic castrate-resistant prostate cancer [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Innovations in Prostate Cancer Research and Treatment; 2026 Jan 20-22; Philadelphia PA. Philadelphia (PA): AACR; Cancer Res 2026;86(2_Suppl):Abstract nr A033.