A Dual-Mode In Situ EV-miRNA Profiling Platform for Machine-Learning-Assisted Gastric Cancer Liquid Biopsy

EV-miRNAs are promising gastric cancer (GC) biomarkers for early diagnosis, yet their clinical application is limited by major detection challenges arising from the low abundance and high sequence homology of EV-miRNAs. Herein, we developed a dual-mode platform for sensitive EV-miRNA in situ profiling based on liposome-encapsulated localized exponential catalytic hairpin assembly (L-LECHA). L-LECHA enables the localized exponential catalytic hairpin assembly system to be delivered into EVs and activates exponential signal amplification, resulting in sensitive and rapid detection of EV-miRNAs. Based on the L-LECHA, this platform achieved a limit of detection of 52.48 aM, representing a 10.96-fold improvement over typical ECHA. Compared to single-mode detection, the dual-mode platform demonstrated a wider linear detection range and greater accuracy. By combining an EV-miRNA panel with a k-nearest neighbors (KNN) model, this platform achieved an AUC of 0.974 for diagnosis of GC. It can also be used for pathological classification of GC. This work provides a robust tool for GC liquid biopsy and precision management.