Abstract Prostate cancer (PCa) is the second most prevalent cancer in men, with early detection and precise staging remaining critical for improving outcomes. While PSA tests and biopsies are widely used, they suffer from low specificity and often lead to unnecessary procedures. Extracellular vesicles (EVs) have emerged as promising noninvasive biomarkers, but their broader application is hindered by the difficulty of isolation and purification from biofluids. Here, we propose an aqueous two‐phase system (ATPS)‐based approach for EV purification to support PCa detection and staging. We isolated plasma EVs (pEVs) from patient plasma using ATPS, a simple and rapid method that requires no specialized equipment. Single EV analysis confirmed efficient removal of plasma impurities and improved accessibility of EV surface biomarker profiling. Machine learning‐based analyses of pEV‐derived markers alongside PSA revealed that the contribution of each marker varies with the diagnostic objective: PSA was most informative for distinguishing benign prostatic hyperplasia (BPH) from cancer cases, while CD81 and PSMA were more informative for tumor staging. Based on these findings, we developed a two‐step machine learning model comprising a screening step (BPH vs. PCa) and a staging step (localized vs. locally advanced PCa). Each step employs selected markers optimized for its diagnostic objective. This step‐optimized sequential model reduces diagnostic errors while using fewer biomarkers, supporting model simplification and clinical scalability. Our approach provides a practical alternative for PCa diagnosis, particularly in settings with limited resources.
Kwon et al. (Thu,) studied this question.