EV-Finder: Direct Detection of Extracellular Vesicle-Associated Proteins by Proximity Extension Assay for Multi-Cancer Screening

Early cancer detection using minimally invasive biomarkers remains a significant challenge, particularly in early-stage disease, where circulating tumor DNA is often below the limit of detection. Extracellular vesicles (EVs), which are actively secreted by viable cancer cells and carry tumor-associated proteins, represent a promising alternative target for liquid biopsy. In this study, we developed EV-finder®, a conceptual framework for the direct detection of EV-associated proteins in serum using proximity extension assay (PEA) technology. Unlike conventional EV-based analytical methods that require prior EV isolation or enrichment, the EV-finder approach enables direct profiling of EV-associated proteins from small serum volumes without an EV isolation step, thereby simplifying the analytical workflow while preserving EV-derived molecular information. Using serum samples from patients with five cancer types (n = 193) and independent healthy controls (n = 138), we established a two-step supervised machine learning framework for cancer detection and tissue-of-origin prediction. The screening model demonstrated promising discriminative performance, with an AUC of 0.985, sensitivity of 0.929, and specificity of 0.957. Notably, no false positives were observed in an external Japanese control cohort, whereas 4 of 29 Korean control samples were classified as cancer-positive. Analysis of EV-associated protein profiles identified both pan-cancer and cancer-type-specific signatures, supporting their value for multi-cancer detection. Collectively, these findings demonstrate the potential feasibility of direct detection of EV-associated proteins from serum using PEA technology and highlight its potential as a scalable and minimally invasive strategy for multi-cancer screening.