Machine learning-assisted dual-amplified visual platform based on fluorescent metal-organic framework and photonic crystals enables ATP detection for pathogen monitoring.

Adenosine triphosphate (ATP), an important energy currency for maintaining life activities, has been widely regarded as a biomarker of disease diagnosis and an indicator of microbial contamination. However, most of the current ATP recognition relies on the specific recognition of aptamer or luciferase, which is prone to degradation and inactivation; and professional equipment is still needed for the signal output. Hence, it still remains challenging to achieve reliable and convenient detection of ATP in an equipment-free way. Herein, we developed a machine learning (ML)-assisted dual-amplified visual platform based on fluorescence metal-organic framework and photonic crystals (PCs) for the visualized and quantitative analysis of free and microbial ATP. Gold nanoclusters (AuNCs)-loaded zeolitic imidazole frameword-8 (ZIF-8) possessed enhanced fluorescence, which could specifically recognize ATP to produce signal variation. Benefiting from PCs-based fluorescence enhancement, visualization and quantification of ATP or E. coli with higher sensitivity could be achieved by capturing fluorescent images and analyzing digital color values. By detecting microbial ATP, E. coli level could be visualized and quantified with a wider linear range of 101-108 CFU mL-1 and a lower LOD of 3.54 CFU mL-1. The validation with E. coli-spiked complicated samples confirmed the feasibility and practical applicability of this sensing platform. Furthermore, five different ML models are designed for the regression quantification of ATP or E. coli with good predictive performance, offering an effective tool for food safety monitoring and clinical diagnosis.