Deep-Learning-Assisted Rational Design of a High-Brightness Dual-Lock Chemiluminescent Probe for Ultrasensitive Detection of Salmonella

Abstract

The development of chemiluminescent probes that simultaneously achieve high brightness and stability remains a significant challenge for ultrasensitive bioassays. Herein, we report a deep-learning-assisted rational design of a novel dual-lock chemiluminescent probe for ultrasensitive Salmonella detection. A fully connected network was trained to predict chemiluminescence intensity, successfully identifying a cyclobutane-acrylonitrile-substituted phenoxy-1,2-dioxetane as an optimal high-brightness chemiluminophore. To circumvent its inherent instability, a dual-lock strategy was implemented: the first lock is opened by a Salmonella-associated C8-esterase activity to accumulate a stable pre-chemiluminophore, and the second lock is subsequently unlocked by in situ-generated singlet oxygen (1O2) via a [2 + 2] cycloaddition, triggering a burst of chemiluminescence. The resultant probe, DSCP-ANCB, exhibits exceptional performance, enabling Salmonella detection with a limit of detection (LOD) as low as 1.9 CFU mL−1 within a total turnaround time of 3 to 8 h, which represents a 1008-fold enhancement in sensitivity over the commercial MUCAP probe. Furthermore, it shows 100% positive and 96% negative agreement with mass spectrometry in clinical isolates (n = 113) and significantly outperforms the commercial fluorescent probe in complex food and environmental matrices. This work introduces a deep-learning-assisted strategy for developing high-performance chemiluminescent diagnostic tools. By integrating computational prediction with rational probe design, this framework provides a generalized approach to improve the overall performance of chemiluminescent probes.