Machine Learning-Empowered Multicolor Detection and Discrimination of Dopaminergic Agents: Utilizing Gold Nanorods for Generating Rainbow Signals

The demand for rapid and sensitive multianalyte detection of dopaminergic agents in assessing the progression of Parkinson’s disease (PD) has attracted recurring interest in recent years. In this study, a multicolor sensor was devised to multiplex detection and discrimination of levodopa (l-DOPA), carbidopa (Carbi), and benserazide (Benz)─key dopaminergic compounds. Notably, the presence of dopaminergic agents inhibits the etching of gold nanorods (AuNRs) by N-bromosuccinimide (NBS), with varying degrees of inhibition depending on the concentration and properties of the analytes. This interaction induces alterations in the size and aspect ratio of the AuNRs, resulting in a red shift within the plasmonic band. Remarkably, this shift in the plasmonic band of AuNRs corresponds to color variations akin to a rainbow in the solution, providing a valuable opportunity for precise visual detection. Statistical analysis of the collected data set was performed using two machine learning techniques: linear discriminant analysis (LDA) for pattern recognition and partial least-squares regression (PLSR) for regression analysis. LDA effectively discriminated dopaminergic agents and their binary and ternary mixtures with 100% accuracy, while PLSR enabled quantitative analysis over a wide concentration range: 1.17–25.0, 2.63–25.0, and 1.18–25.0 μmol L–1, with detection limits of 0.93, 0.88, and 0.39 μmol L–1, for l-DOPA, Carbi, and Benz, respectively. Furthermore, we validated the practical applicability of the sensor assay by identifying and quantifying dopaminergic agent in medicines tablets containing mixtures of these compounds, affirming its utility and reliability for on-site medicines quality control.