Machine Learning-Driven 3d-Qsar Models Facilitated Rapid On-Site Broad-Spectrum Immunoassay of (Fluoro)Quinolones Using Evanescent Wave Fiber-Embedded Optofluidic Biochip

Abstract

(Fluoro)quinolones (FQs) pose significant threats to public health due to their widespread use and persistence in food and water sources. Given the extensive variety of FQs, testing each compound individually is prohibitively expensive and time-consuming. Here, we introduce an evanescent wave fiber-embedded 3D optofluidic biochip (e-FOB) that enables rapid, on-site detection of 14 FQs through a broad-spectrum immunoassay. The e-FOB integrates a functionalized fiber biosensor with a 3D optofluidic chip, leveraging a broad-spectrum anti-FQ antibody to achieve high sensitivity and specificity. Limits of detection for all tested FQs were below 3.0 µg/L, with excellent reusability and stability demonstrated over 400 cycles. Two machine learning-driven 3D quantitative structure-activity relationship (ML-3D-QSAR) models were developed to identify key physicochemical factors and quantitative interactions influencing FQs detection performance. Two ML-3D-QSAR models, CoMFA and CoMSIA, can accurately predict the detection performance of other unknown FQs, highlighting the potential for broad-spectrum detection of antibiotics. The e-FOB was successfully applied to detect FQs in complex matrices such as honey and water samples, demonstrating its practical applicability. The ML-3D-QSAR empowered e-FOB offers a revolutionary approach to rapid, on-site screening of antibiotic residues, improving detection efficiency and reducing costs while protecting public health.

Keywords: Fiber-embedded 3D optofluidic biochip, (Fluoro)quinolones, Broad-spectrum antibody, Evanescent wave fluorescence, machine learning, Quantitative structure-activity relationship