Early detection of renal cell carcinoma: A novel cfDNA fragmentomics-based liquid biopsy assay.

Abstract

4534 Background: Renal cell carcinoma (RCC) is a leading cause of cancer-related mortality, with a rapidly rising global incidence. Early detection greatly improves the outcomes of RCC, yet current diagnostic methods have limitations in sensitivity, specificity, and accessibility. This study develops and evaluates a cfDNA fragmentomics-based liquid biopsy integrated with machine learning as a non-invasive and scalable tool for RCC early detection. Methods: This case-control cohort study recruited 442 participants (223 RCC patients and 219 non-cancer controls, including healthy individuals and those with benign renal conditions) at a single cancer referral center from December 2021 to December 2023. Plasma-derived cfDNA underwent low-pass WGS (5X coverage), and three fragmentomics features—copy number variation (CNV), fragment size ratio (FSR), and nucleosome footprint (NFP)—were extracted. A stacked ensemble machine learning model was trained on 280 participants and validated on 162 independent participants. Performance was assessed using area under the curve (AUC), sensitivity, and specificity. Results: The ensemble model achieved an AUC of 0.9656 in the validation cohort, with sensitivity and specificity of 90.5% and 93.8%, respectively. Stratified analyses demonstrated consistent performance across RCC stages, histological subtypes, and Fuhrman grades, with sensitivities of 87.8% for Stage I RCC and 100% for Stage IV RCC. Additionally, the model effectively differentiated malignant RCC from benign renal conditions, further validating its clinical utility. Stability evaluations confirmed the model's robustness across diverse sample types, storage conditions, and processing scenarios, underscoring its potential applicability in routine clinical practice. Conclusions: This cfDNA fragmentomics-based liquid biopsy represents a highly sensitive, specific, and non-invasive approach for the early detection of RCC. Its robust performance across diverse clinical scenarios highlights its potential to enhance current RCC diagnostic workflows, facilitate timely interventions, and improve patient outcomes. Future integration of this method into clinical practice could address critical gaps in RCC management, providing substantial benefits in early detection and personalized care.