Machine learning-powered plasmonic pattern recognition: etch-suppressed gold nanorods for multiplex urinary analysis of catecholamine neurotransmitters

Abstract

Simultaneous monitoring of catecholamine neurotransmitters (CNTs)—including epinephrine (Epi), norepinephrine (NE), levodopa (L-DOPA), and dopamine (DA)—is essential for the accurate diagnosis and effective management of various neurological disorders. However, most existing sensing platforms face challenges such as limited multiplexing capabilities and potential cytotoxicity of the sensing components. In this work, we introduce a sensitive, non-toxic, and non-invasive single-component multicolorimetric probe capable of detecting and distinguishing low concentrations of Epi, NE, L-DOPA, and DA, along with their binary, ternary, and quaternary mixtures. The sensing mechanism relies on the controlled inhibition of gold nanorod (AuNR) etching in the presence of varying concentrations of N-bromosuccinimide (NBS), a mild oxidizing agent, under ambient conditions. To enable robust analyte identification and quantification, the spectral responses were processed using machine learning algorithms—specifically, linear discriminant analysis (LDA) for classification and partial least squares regression (PLSR) for concentration prediction. Under the optimized conditions, the assay demonstrated excellent linearity across a broad concentration range for each analyte, Epi (1.8–20 μmol L⁻¹), NE (1.4–25 μmol L⁻¹), L-DOPA (1.5–25 μmol L⁻¹), and DA (2.8–50 μmol L⁻¹), with corresponding detection limits of 0.62, 0.47, 0.49, and 0.92 μmol L⁻¹. The practical utility of the platform was validated through successful application to human urine samples, confirming its potential as a powerful tool for point-of-care diagnostics and clinical neurochemical analysis.