Deep Learning‐Enhanced DNAzyme‐Driven Rolling‐Circle Amplification Encoding for Multibacterial Detection

ABSTRACT Food‐borne outbreaks are frequently caused by multiple live pathogens that conventional methods cannot process simultaneously. We report a DNAzyme‐driven rolling‐circle amplification/molecular‐beacon encoding system (DRM‐ES) coupled with a smartphone‐based convolutional neural network (CNN) that simultaneously identifies and quantifies three live bacteria from 32 real‐world samples. Bacteria‐secreted proteins cleave bead‐immobilized DNAzymes, releasing primers that initiate RCA and generate long concatemers; each opens a spectrally distinct molecular beacon, producing blue, green, or red fluorescence captured in one smartphone image and decoded by a CNN trained on 2800 images. DRM‐ES achieves 10 1 –10 2 CFU/mL sensitivity for S. aureus , B. cocovenenans , and E. coli in food, clinical, and environmental samples; shows 100% positive and ≥95.2% negative agreement with culture; and correctly identifies 29/32 samples naturally contaminated with these three bacteria in a 32‐tube array. The platform offers culture‐comparable sensitivity and live‐cell specificity, providing a generalizable blueprint for large‐scale multiplex pathogen screening.