PruEV-AI: a Simple Approach Combines Urinary Extracellular Vesicle Isolation with AI-Assisted Analysis for Prostate Cancer Diagnosis.

Abstract

Urinary extracellular vesicles (uEVs) are a promising source of prostate-derived biomarkers for non-invasive prostate cancer (PCa) diagnosis. However, conventional uEV isolation methods and single-marker assays often lack efficiency and diagnostic accuracy. Here, PruEV-AI is introduced, an integrated diagnostic system that combines rapid uEV isolation with AI-based biomarker analysis. The PruEV platform employs amine-modified zeolites (AZ) and carbohydrazide (CDH) to isolate uEVs and extract miRNAs in less than 30 min through electrostatic and covalent interactions. This one-step syringe-filter process enables high-throughput, reproducible, and user-friendly isolation of uEVs suitable for clinical diagnostics. Among 12 candidate miRNAs, 6 are validated using RT-qPCR in urine samples from 48 PCa patients and 49 controls. Individually, these miRNAs and PSA show modest diagnostic performance, with area under the curve (AUC) values ranging from 0.6 to 0.8. To overcome the limitations of single biomarkers, a deep learning (DL) model evaluates all 127 possible combinations of the 6 miRNAs and PSA. The optimal biomarker combination identified by the DL model achieves an AUC of 0.9556, with 93.33% sensitivity, specificity, and accuracy. Consequently, the PruEV-AI system provides a robust, non-invasive, and clinically relevant diagnostic approach for accurately identifying PCa, thereby supporting improved screening protocols and more effective therapeutic strategies.