Reactive silver ink-derived SERS-active nanofiber platform enabling machine learning-assisted sensitive amino acid...

Surface-Enhanced Raman Spectroscopy (SERS) has long been a recognized method for the rapid and sensitive detection of low concentrations of analytes. Nevertheless, challenges still exist in the fabrication of practical, effective SERS substrates and the interpretation of complex SERS spectra. In this study, a machine learning (ML)-assisted SERS platform that is both scalable and practical is proposed. The SERS substrate, composed of silver nanoparticle-decorated nanofibers (AgNFs), was obtained by integrating nanoparticle-free reactive silver (Ag) ink into a nanofiber matrix containing polyvinylpyrrolidone (PVP) and poly (ethylene oxide) (PEO). The nanofibers served both as a mechanical scaffold and a nano-scale template, guiding the formation of homogeneously distributed Ag nanostructures throughout the substrate. This synergistic fiber-assisted growth produced a highly SERS-active substrate, resulting in an analytical enhancement factor (AEF) of 2.9 × 10⁷. This platform enabled detection of phenylalanine, proline, valine, alanine, and cysteine amino acids at concentrations as low as 0.3 μM. The results demonstrated a direct correlation between the analyte concentration and the SERS signal intensity, as evidenced by determination coefficients (R²) that exceeded 0.9. Furthermore, the spectral data were subjected to analysis using ML algorithms, thereby allowing for the classification of amino acids. The linear support vector machine (SVM) algorithm exhibited superior classification performance, with an accuracy rate of approximately 98%. The findings demonstrate the efficacy of the developed ML-enhanced SERS platform for detecting various analytes, as well as its advantageous properties of simplicity, scalability, reproducibility, and ultra-sensitive structure.