Quantitative multiplexing of uric acid and creatinine using polydisperse plasmonic nanoparticles enabled by electrochemical SERS and machine learning

Abstract

Surface-enhanced Raman spectroscopy (SERS) is a promising technique for the detection of biomarkers, but it can struggle to quantify multiple analytes in complex fluids. This study combines electrochemical SERS (E-SERS) and machine learning for the quantitative multiplexed detection of uric acid (UA) and creatinine (CRN). Using classical polydisperse Ag nanoparticles (NPs) made by scalable synthesis, we achieved quantitative multiplexing with low limits of detection (LoDs) and high prediction accuracy, comparable to those made by sophisticated approaches. The E-SERS LoDs at the optimal applied potentials were 0.127 μM and 0.354 μM for UA and CRN respectively, compared to 0.504 μM and 1.02 μM for conventional SERS (recorded at 0V). By collecting a multi-dimensional E-SERS dataset and applying a two-step partial least squares regression – multilayer perceptron (PLSR-MLP) machine learning algorithm, we were able to identify the analyte concentrations in unseen spectra with a prediction accuracy of 0.94. This research demonstrates the potential of E-SERS and machine learning for multiplexed detection in clinical settings.