Prediction of Prostate Cancer Biochemical Recurrence After Radical Prostatectomy by Collagen Models Using Multiomic Profiles

Abstract

Background and objective: The interplay between prostate cancer and the tumor microenvironment is well documented and of primary importance in disease evolution. Herein, we investigated the prognostic value of tissue and urinary collagen-related molecular signatures in predicting biochemical recurrence (BCR) after radical prostatectomy (RP).

Methods: A comprehensive analysis of 55 collagen-related features was conducted using transcriptomic datasets (n = 1393), with further validation at the proteomic level (n = 69). Additionally, a distinct cohort (n = 73) underwent a urine-based peptidomic analysis, culminating in the validation of a urine-derived prognostic model. Independent prognostic significance was assessed using Cox proportional hazards modeling, while the model's predictive performance was benchmarked against established clinical metrics.

Key findings and limitations: An expression analysis of 55 collagen-related transcripts identified 11 transcripts significantly associated with BCR (C-index: 0.55-0.72, p < 0.002). Multivariable models incorporating these transcripts enhanced prognostic accuracy, surpassing clinical variables (C-index: 0.66-0.89, p < 0.002). Proteomic validation confirmed five key collagen proteins, while a urine-based collagen model (C-index: 0.73, 95% confidence interval: 0.62-0.85) demonstrated a strong prognostic potential, although limited by small patient numbers. Additionally, tissue collagen-based models predicted overall survival with a significant prognostic value (C-index: 0.59-0.70, p < 0.01).

Conclusions and clinical implications: Collagen-based molecular signatures in both tissue and urine emerge as robust prognostic biomarkers for predicting BCR following RP.

Keywords: Collagen metabolism; Machine learning; Postprostatectomy biochemical recurrence; Predictive biomarker; Prostate cancer.