We present a simple, fast, and efficient method for the production of silver nanoparticles (AgNP) for surface-enhanced Raman spectroscopy (SERS) applications. The method utilizes the classical silver mirror reaction, with ascorbic acid serving as the reducing agent, to produce AgNP on polyacrylonitrile non-woven substrates. Different protocols were used to functionalize the substrates with AgNP in situ. Optimization of the substrate modification procedure led to SERS-substrates with an enhancement factor (EF) of up to 10⁶ (4-MBA detection) for samples treated with 1 M solutions of AgNO₃ and NH₃·H₂O. The obtained substrates were further modified with glucose oxidase (GOx), and glucose sensing was carried out in a concentration range from 0.5 mM to 10 mM. Machine learning models, such as random forest and gradient boosting, were used for the analysis of Raman data. The resulting substrates have potential applications in various fields of analytical chemistry.

In recent years, non-invasive methods for the analysis of biological fluids have attracted growing interest. In this study, we propose a straightforward approach to fabricating silver nanoparticle (AgNP)-coated non-woven polyacrylonitrile substrates for surface-enhanced Raman scattering (SERS). AgNPs were synthesized directly on the substrate using borohydride reduction, ensuring uniform distribution. The optimized SERS substrates exhibited a high enhancement factor (EF) of up to 10⁵ for the detection of 4-mercaptobenzoic acid (4-MBA). To enable glucose sensing, the substrates were further functionalized with glucose oxidase (GOx), allowing detection in the 1–10 mM range. Machine learning classification and regression models based on gradient boosting were employed to analyze SERS spectra, enhancing the accuracy of quantitative predictions (R² = 0.971, accuracy = 0.938, limit of detection = 0.66 mM). These results highlight the potential of AgNP-modified substrates for reliable and reusable biochemical sensing applications.