Abstract
The escalating global burden of infectious diseases demands biosensing technologies that transcend the complexity-sensitivity-accuracy trade-off in real-world applications. Herein, an interpretable machine learning-empowered multimodal biosensor synergizing electron transfer-enhanced nanozymes and aggregation-induced emission luminogens (AIEgens) for ultrasensitive pathogen detection is presented. By engineering aminophenol formaldehyde resin nanobowls anchored with monodisperse Pt nanoparticles, interfacial electron transfer (N→Pt→O) induces an upshift of Pt d-band center relative to the Fermi level, as validated by density functional theory. This electronic modulation optimizes H(2)O(2) adsorption energy, lowers the energy barrier of the rate-determining step, and reduces activation energy, resulting in a 3.4-fold enhancement in peroxidase-like activity over conventional Pt nanozymes. Then, AIEgens are strategically integrated to generate cross-validated anti-interference signals, achieving a record-low detection limit for Salmonella typhimurium, surpassing classical immunoassays in sensitivity and accuracy. A SHapley Additive exPlanations (SHAP)-guided eXtreme Gradient Boosting (XGBoost) algorithm dynamically fuses multimodal signals, enhancing sensitivity by five fold over single-mode detection and delivering 100% diagnostic accuracy for positive samples. SHAP further deciphers the synergetic mechanism, revealing concentration-dependent signal contributions and validating decision logic. This work pioneers a nanozyme-AI co-design framework, bridging d-band-driven catalytic precision and machine learning-powered signal intelligence to redefine biosensing paradigms for combating public health emergencies.