Municipal wastewater sludge/laser-induced graphene nanozyme for machine-learning-assisted smartphone colorimetric tetracycline sensing

Municipal wastewater sludge (MWS) is an abundant Fe-containing waste stream with significant potential for valorization as a functional catalytic material. Herein, an Fe-rich MWS-derived phase was integrated with polyimide-derived laser-induced graphene (LIG) to construct a multifunctional LIG/MWS composite nanozyme for tetracycline (TC) sensing. LIG was fabricated directly using a low-cost 450 nm diode laser, while MWS was recovered through coagulation-flocculation, calcination, acid washing, and drying. The optimized LIG/MWS composite exhibited enhanced peroxidase-like activity toward the TMB-H₂O₂ system, where TC inhibited color development through interaction with Fe-containing catalytic sites. The proposed colorimetric assay enabled TC determination over 0-50 µM with a detection limit of 0.70 µM and R ² = 0.9866. The sensing strategy was further translated into a smartphone-readable paper-based analytical device (PAD), where RGB-based calibration provided portable TC screening capability. In parallel, a separate 96-well smartphone-imaging dataset was analyzed using principal component analysis (PCA) and supervised machine learning to evaluate multivariate concentration prediction and acquisition-condition variability. PCA revealed that the first two principal components explained 97.10% of RGB variance, while illumination was identified as the dominant source of systematic variation. Among PLSR, SVR, random forest, XGBoost, and ANN models, ANN achieved the best nested grouped-cross-validation performance (R ² = 0.987, RMSE = 6.95 µM), whereas leave-one-condition-out validation highlighted outdoor illumination as the major limitation for model generalization. The LIG/MWS composite also demonstrated secondary methylene blue adsorption, highlighting its multifunctional potential. This work provides a sustainable nanozyme platform integrating waste valorization, smartphone-assisted sensing, and machine-learning-enabled colorimetric analysis.