Machine learning-assisted electrochemical SERS for sensitive detection of multiple urinary proteins

In this study, we report the development and application of a plasmonic nanostructure-based surface-enhanced Raman spectroscopy (SERS) platform, integrated with electrochemical deposition and a convolutional neural network (CNN), for the sensitive and precise detection and quantification of kidney disease biomarkers in urine. A gold nanodimple (AuND) substrate was subjected to electrochemical deposition to produce a highly reproducible and sensitive SERS substrate for protein analysis. Evaluation of the performance of the platform through reproducibility tests and quantitative analysis of Cytochrome C protein demonstrated that it has a high sensitivity and a low detection limit of 8.2 pg/mL. The system was then applied to multiplexed analysis of three kidney disease-related proteins: albumin, transferrin (TrF), and immunoglobulin G (IgG). A combination of ANOVA-based feature selection and CNN classification models achieved high accuracy, with classification accuracies of 93.8% for albumin, 96.8% for TrF, and 96.8% for IgG. Subsequently, CNN-based regression models were utilized to quantify protein concentrations in urine samples, demonstrating robust performance with R 2 values of 0.9321 for albumin, 0.9848 for TrF, and 0.9957 for IgG. The method also exhibited excellent diagnostic feasibility, successfully detecting and quantifying target proteins in a urine matrix. The proposed platform thus offers a highly sensitive, reliable and non-invasive approach for early diagnosis of kidney diseases.