ESMO Open. 2025 Jun 20. pii: S2059-7029(25)01192-5. [Epub ahead of print]10(7): 105323
Y L Peng,
B Yu,
T X Huang,
Z H Zhou,
H Zhang,
W X F Tang,
X X Xu,
D Q Zhu,
R W Yang,
H Bao,
X Wu,
H Han,
Zh L Zhang,
L R He,
P Dong,
W S Wei.
BACKGROUND: Renal cell carcinoma (RCC) is a growing global health challenge, with poor survival rates in advanced stages due to the lack of effective early detection methods. This study developed a cell-free DNA (cfDNA) fragmentomics-based liquid biopsy to enable noninvasive RCC diagnosis.
PATIENTS AND METHODS: Using low-pass whole-genome sequencing, three cfDNA features-copy number variation, fragment size distribution, and nucleosome footprint-were derived and integrated into a stacked ensemble machine learning model. A total of 280 participants, including 142 RCC patients and 138 noncancer controls, were enrolled for model development. The resultant model was subsequently evaluated using both validation and external cohorts.
RESULTS: The developed model achieved an area under the curve of 0.966 and 0.952 in the validation and external cohorts, respectively. In the validation cohort, the model demonstrated a sensitivity of 90.5% and specificity of 93.8%, while in the external cohort, it exhibited a sensitivity of 76.7% and specificity of 92.9%. Robust performance was observed across different TNM (tumor-node-metastasis) stages, histological subtypes, and Fuhrman grades, with high sensitivity even for early-stage RCC. Furthermore, the assay effectively differentiated malignant tumors from benign renal conditions, thereby potentially reducing unnecessary surgical interventions.
CONCLUSION: The cfDNA fragmentomics-based liquid biopsy provides a promising, noninvasive, cost-effective, and accurate approach for RCC detection and management, substantiating its potential as a complementary tool to conventional diagnostic techniques. Future research should explore integrating cfDNA fragmentomics with multi-omics approaches to enhance diagnostic precision and broaden clinical utility.
Keywords: circulating tumor DNA; early cancer detection; fragmentomics; liquid biopsy; machine learning; renal cell carcinoma