Machine Learning-Aided Identification of Fecal Extracellular Vesicle microRNA Signatures for Noninvasive Detection of Colorectal Cancer

Colorectal cancer (CRC) remains a formidable threat to human health, with considerable challenges persisting in its diagnosis, particularly during the early stages of the malignancy. In this study, we elucidated that fecal extracellular vesicle microRNA signatures (FEVOR) could serve as potent noninvasive CRC biomarkers. FEVOR was first revealed by miRNA sequencing, followed by the construction of a CRISPR/Cas13a-based detection platform to interrogate FEVOR expression across a diverse spectrum of clinical cohorts. Machine learning-driven models were subsequently developed within the realms of CRC diagnostics, prognostics, and early warning systems. In a cohort of 38 CRC patients, our diagnostic model achieved an outstanding accuracy of 97.4% (37/38), successfully identifying 37 of 38 CRC cases. This performance significantly outpaced the diagnostic efficacy of two clinically established biomarkers, CEA and CA19-9, which showed accuracies of mere 26.3% (10/38) and 7.9% (3/38), respectively. We also examined the expression levels of FEVOR in several CRC patients both before and after surgery, as well as in patients with colorectal adenomas (CA). Impressively, the results showed that FEVOR could serve as a robust prognostic indicator for CRC and a potential predictor for CA. This endeavor aimed to harness the predictive power of FEVOR for enhancing the precision and efficacy of CRC management paradigms. We envision that these findings will propel both foundational and preclinical research on CRC, as well as clinical studies.