Detection of adenovirus using dendritic silver substrates combined with SERS and random forest algorithms

• Published in: Applied surface science 2025-02, Vol.682, p.161771, Article 161771
• Main Authors: Tabarov, Artem, Prigoda, Kristina, Popov, Evgeniy, Ermina, Anna, Levitskii, Vladimir, Krylov, Danila, Andreeva, Olga, Gazizulin, Azat, Bolshakov, Vladimir, Tolmachev, Vladimir, Markov, Danila, Amosova, Irina, Timoshicheva, Tatyana, Gorshkov, Andrey, Danilenko, Daria, Vitkin, Vladimir, Zharova, Yuliya
• Format: Article
• Language: English
• Subjects: Adenovirus; Machine learning; Random forest algorithm; SAXS; SERS; Silver dendrites
• ISSN: 0169-4332
• DOI: 10.1016/j.apsusc.2024.161771

[Display omitted] •SERS substrates based on Ag dendrites were chemically synthesized and compared with a commercial “SERSitive” substrate.•Adenovirus detection was performed using SERS in combination with a machine learning (random forest algorithm).•Classification accuracy ranged from 72 to 92 %.•Particle size of adenovirus (82 ± 2 nm) was measured by a small angle X-ray scattering technique. In this study, three surface-enhanced Raman scattering (SERS) substrates based on silver dendrites with heights of ∼300, 400, and 600 nm were chemically synthesized on silicon wafer and compared with a commercial SERS (SERSitive, Poland) substrate. Optical investigation of Ag dendrites revealed the manifestation of localized surface plasmon resonance (LSPR) in the spectral region of 690 nm. This spectral band of the LSPR is close to the wavelength of the exciting He-Ne laser used in this work, which enhances the Raman signal. Adenovirus samples were measured by a small angle X-ray scattering technique which determined the sizes (82 ± 2 nm) and shapes (icosahedron) of the viral particles. Adenovirus detection was performed using SERS in combination with machine learning algorithms (random forest algorithm). The study concluded that dendritic SERS substrates demonstrate potential for adenovirus detection and diagnostics comparable to the SERSitive substrate. The classification of spectra from pure buffer solutions and buffer solutions with adenovirus was achieved with classification accuracy ranging from 72 to 92 %.