A Digital Score Based on Circulating-Tumor-Cells-Derived mRNA Quantification and Machine Learning for Early Colorectal Cancer Detection

Abstract

Circulating tumor cells (CTCs) serve as valuable biomarkers in tumor circulation, carrying essential primary tumor information. The purification of CTCs from peripheral blood samples and the analysis of their characteristic molecules enable the detection of tumors at an early stage. The noninvasive, continuous, real-time dynamic monitoring provides a promising solution for the timely diagnosis of colorectal cancer (CRC). In this study, we developed a minimally invasive method for CRC early detection to enable accurate screening in a friendly manner for individuals who generally require colonoscopy. The dual-antibody (i.e., anti-EpCAM and anti-EGFR) modified antifouling hydrogel-coated magnetic nanoparticles (pSBMA-MNPs) were prepared for efficient and specific CTC purification. Then, the quantification of 6 RNA transcripts in purified CRC CTCs was performed via droplet digital PCR (ddPCR), and a CRC score was calculated using an extreme gradient boosting model to distinguish CRC from colon polyps and adenomas. A pilot study was conducted to evaluate the clinical potential of the CRC CTC RNA assay in a training cohort ( n = 101) and an independent test cohort ( n = 65), achieving a diagnostic accuracy of 91.0% in the whole cohort, significantly outperforming serum CEA, CA125, and CA199. Subgroup analysis across CRC stage, age, and tumor location of patients was also performed, and the CRC score exhibited robust performance, demonstrating commendable diagnostic efficacy for CRC detection and promising application in friendly screening individuals that really require colonoscopy.