Machine Learning-Enhanced Glucose Sensing via Dual-Synthesis AgNP SERS Substrates: From High-Throughput Fabrication to Clinical Detection
Ekaterina Prikhozhdenko1, Viktoriia Bakal2, Olga Gusliakova2, Anastasia Kartashova2, Mariia Saveleva2, Valentina Plastun2, Polina Demina2, Ilya Kozhevnikov2, Evgenii Ryabov2, Daniil Bratashov1,2, Alexey Serdobintsev2
1 Moscow Institute of Physics and Technology, Dolgoprudny, Moscow region, Russia
2 Saratov State University, Saratov, Russia
Abstract
Two scalable approaches for fabricating reusable silver nanoparticle (AgNP)-functionalized polyacrylonitrile substrates for surface-enhanced Raman scattering (SERS) biosensing are presented. The first strategy employs in-situ borohydride reduction (via direct synthesis) to achieve uniformly distributed AgNPs, yielding substrates with enhancement factors (EF) of 10⁵ for 4-MBA detection. The second method utilizes an optimized silver mirror reaction with ascorbic acid (through refined protocols), achieving higher EFs up to 10⁶ using 1M AgNO₃/NH₃·H₂O. Both platforms were functionalized with glucose oxidase (GOx) for non-invasive glucose detection, covering clinically relevant ranges (0.5–10 mM) in biological fluids.
Critical innovation lies in applying machine learning to overcome spectral complexity: Gradient boosting models achieved quantitative glucose prediction (R²=0.971) with 93.8% accuracy and 0.66 mM detection limit in the borohydride synthesis method, overcoming limitations of conventional spectral analysis. Substrates from the silver mirror reaction leveraged both random forest and gradient boosting for robust analysis. The AgNP substrates demonstrated exceptional reusability (>10 cycles) and batch reproducibility (RSD <8%), highlighting their practicality for point-of-care use.
These integrated results establish that:
(i) Both synthesis routes enable cost-effective, high-throughput SERS substrate production;
(ii) ML-driven analytics significantly enhance prediction accuracy beyond traditional methods;
(iii) The platforms show clinical utility for glucose monitoring in physiological ranges relevant to diabetes management. The methodologies are extensible to diverse biomarkers, opening avenues for real-time point-of-care diagnostic devices.
Speaker
Ekaterina S. Prikhozhdenko
Laboratory of medical equipment in the field of in vitro diagnostics, Moscow Institute of Physics and Technology, Dolgoprudny, Moscow region, Russia
Россия
Discussion
Ask question
