A machine learning-assisted colorimetric platform based on a Cu-doped carbon dot nanozyme for sensitive detection of N-acetylcysteine in serum.

Abstract

N-Acetylcysteine (NAC), a thiol-containing mucolytic agent for COPD, exhibits significant pharmacokinetic variability among patients, motivating the need for convenient therapeutic monitoring. Herein, a colorimetric platform based on copper-doped chiral carbon dots (Cu-D-CDs) was synthesized via a one-pot hydrothermal route using D-histidine and CuCl2 as precursors. The Cu-D-CDs displayed enhanced peroxidase-like activity, catalyzing H2O2-mediated oxidation of TMB to blue oxTMB (λmax = 652 nm). Upon NAC introduction, oxTMB was reduced via a thiol-disulfide redox reaction, producing an absorbance decrease proportional to NAC concentration. Under optimized conditions, the assay exhibited a linear dynamic range of 10-90 µM, a detection limit of 3.74 µM (LOD = 3σ/S), and recoveries of 96.00-106.60% (RSD < 4%) in mouse serum, demonstrating its feasibility for NAC quantification in serum samples. Beyond the single-wavelength readout at 652 nm, the full-spectral data were further processed by an LSTM network, which significantly improved prediction accuracy (R2 > 0.9998) compared with single-wavelength calibration (R2 = 0.9979), effectively mitigating matrix interference. This integrated colorimetric-LSTM strategy shows promise for COPD therapeutic drug monitoring and pharmaceutical quality control.