Conventional pathogen detection methods tend to suffer from limitations such as prolonged processing time, operational complexity, or insufficient sensitivity. To address the need for rapid and highly sensitive detection technologies, researchers from China's Hefei University of Technology have developed a machine learning-assisted fluorescent sensor array strategy, constructing a 3 × 6 sensing platform utilizing three water-soluble perovskite quantum dots (PQDs) with distinct fluorescent properties.
The array generates significant fluorescence color changes through electrostatic interactions between PQDs and bacterial surfaces, as well as Aggregation-Caused Quenching (ACQ) effects. Relative fluorescence color changes (ΔRGB) were captured using a smartphone and subsequently analyzed through machine learning algorithms, including K-Nearest Neighbors (KNN) and principal component analysis (PCA).
This methodology facilitated the efficient identification and precise discrimination of six pathogenic bacteria, as well as their binary and ternary mixtures, achieving a limit of detection (LOD) ranging from 92 to 121 CFU mL−1 for individual bacterial species.
In this work, the water-soluble PQDs were synthesized at room temperature through surface modification using the fluorocarbon reagent. Subsequently, by leveraging halide ion exchange reactions and modulating the chloride ion (Cl−) content via the addition of hydrochloric acid (HCl), the fluorescence emission peaks of the PQDs were precisely tuned to generate three distinct types of quantum dots with unique fluorescent properties (GPQD, CPQD, and BPQD).
Based on the aggregation-induced quenching (ACQ) effect, which originates from the electrostatic interactions between PQDs and the negatively charged bacterial surfaces, the 3 × 6 fluorescent sensor array was developed on a microplate. The relative fluorescence color changes (ΔRGB) induced by bacterial-PQDs interactions were digitally recorded via a smartphone-based Color Grab platform. The acquired data were subsequently processed and analyzed using the K-Nearest Neighbors (KNN) algorithm and principal component analysis (PCA) implemented in MATLAB, ultimately enabling efficient recognition and precise identification of multiple target pathogens.
The proposed approach is characterized by its simplicity, cost-effectiveness, and rapidity, demonstrating 100 % accuracy in both blind and real-sample tests. This innovative strategy presents a promising approach to rapid pathogen detection, with potential applications spanning food security, clinical diagnostics, and environmental surveillance.
