Deep Learning-Assisted SERS for Therapeutic Drug Monitoring of Clozapine in Serum on Plasmonic Metasurfaces

Abstract

Clozapine is widely regarded as one of the most effective therapeutics for treatment-resistant schizophrenia. Despite its proven efficacy, the therapeutic use of clozapine is complicated by its narrow therapeutic index, which necessitates rapid and precise therapeutic drug monitoring (TDM) to optimize patient outcomes and minimize adverse effects. However, conventional techniques, such as high-performance liquid chromatography, are limited by their high costs, complex instrumentation, and long turnaround times. Herein, we propose a novel approach that integrates artificial neural networks (ANNs) with surface-enhanced Raman spectroscopy (SERS) on a plasmonic metasurface for rapid TDM of clozapine and its two primary metabolites, norclozapine and clozapine-N-oxide, in human serum. The ANN-SERS strategy enables accurate classification and robust concentration prediction of the three analytes. We envision that the integrated ANN-SERS framework could deliver a scalable biomedical diagnostic and therapeutic tool for studying a wide variety of chemical and biological molecules in clinical settings.