SERS and supervised learning for proof-of-concept rotavirus detection in wastewater matrices

Rotavirus is one of the leading causes of acute gastroenteritis worldwide, and its surveillance in wastewater represents a valuable tool for early outbreak detection and public health monitoring. In this work, we address the need for a rapid, sensitive, and cost-effective sensing platform capable of detecting Rotavirus in complex wastewater matrices without the use of time-consuming molecular amplification techniques. To this aim, we developed a surface-enhanced Raman scattering (SERS) biosensor based on large-area plasmonic substrates fabricated via liquid-liquid self-assembly of monodisperse gold nanoparticles driven by the Marangoni effect. The substrates were functionalized with monoclonal anti-Rotavirus antibodies, and SERS spectra were acquired from both controlled samples and real wastewater matrices. Spectral data were subsequently processed and classified using a supervised machine learning approach based on a support vector machine (SVM). The SERS substrates exhibited high uniformity, reproducibility, long-term stability, and low fabrication cost. The antibody-functionalized biosensor enabled sensitive and selective detection of Rotavirus, with a limit of detection of 3.9 TCID₅₀/mL in phosphate-buffered saline and a clear dose-dependent response. When applied to wastewater matrices, the trained SERS-SVM classification pipeline achieved an average accuracy of 0.89 and a Cohen's κ of 0.79, supporting the feasibility of discriminating between Rotavirus-positive and negative samples. These results highlight the proof-of-concept value of the proposed SERS-machine-learning platform as a screening tool for wastewater-based epidemiology.