Résumé
Background
Early detection of colorectal cancer (CRC) remains a major unmet need in precision oncology. Circulating tumor DNA (ctDNA) has limited sensitivity in early-stage disease because it depends on tumor burden and active shedding. Marketed ctDNA tests report 45–55% sensitivity for stage I CRC, underscoring the need for biomarkers that capture earlier molecular changes. Epitranscriptomics explores dynamic RNA chemical modifications that regulate RNA stability, translation and stress responses. These marks reflect adaptive programs involved in cancer initiation and progression. Unlike genomic alterations, they can reveal functional perturbations of tissue homeostasis at tumor onset, making them relevant for early detection.
Methods
We developed a platform combining high-resolution mass spectrometry and machine learning to quantify circulating modified nucleosides from less than 10 μL of plasma. As these nucleosides cannot be recycled and are excreted into biofluids, they provide a stable, minimally invasive readout of cancer-associated RNA modification programs.
Results
A five-nucleoside circulating epitranscriptomic signature for CRC was identified in 130 CRC patients, stages I–IV, and 30 healthy donors. Using supervised machine learning, it discriminated CRC samples with high accuracy, reaching a cross-validation F1-score ≥ 0.95, including stage I disease. Validation in 47 CRC patients, stages I–IV, and 20 healthy donors confirmed robustness, with precision, recall and F1-score all ≥ 0.95. Notably, discrimination was lost after tumor resection in 20 resectable stage I–II CRC patients, supporting a direct association with tumor presence and suggesting potential to outperform ctDNA benchmarks for early-stage CRC detection. In 64 polyp patients and 54 healthy donors, circulating epitranscriptomic profiles clearly distinguished pre-cancerous lesions, from healthy samples.
Conclusions
Collectively, these findings establish RNA modification profiling as a powerful liquid biopsy modality for CRC detection and management, paving the way for broader clinical applications of epitranscriptomic biomarkers beyond conventional DNA-based approaches.