Machine Learning Integrated Near-Infrared Surface-Enhanced Raman Spectroscopy for Accurate Strain-Level Virus...

Area of Science:

• Nanotechnology
• Spectroscopy
• Artificial Intelligence

Background:

• Strain-level virus identification is crucial for public health but current methods lack speed and sensitivity.
• Surface-enhanced Raman spectroscopy (SERS) offers potential but faces limitations with clinical samples and near-infrared (NIR) excitation.
• Existing diagnostic tools struggle with spectral complexity and generalizability in artificial intelligence (AI) applications.

Purpose of the Study:

• To develop an AI-empowered NIR-SERS platform for rapid and accurate virus identification.
• To overcome limitations of conventional SERS and AI diagnostics in complex clinical samples.
• To enable precise classification of viruses, including challenging strain-level differentiation.

Main Methods:

• Integration of machine learning with a hybrid substrate of gold nanostars (AuNSt) and gold-coated carbon nanotube arrays (AuCNT).
• Utilizing a rationally designed substrate to generate localized plasmonic hot spots resonant to NIR excitation.
• Employing electron energy-loss spectroscopy (EELS) to confirm substrate performance and signal amplification.

Main Results:

• The platform achieved accurate classification of respiratory viruses (influenza and coronaviruses) at type, subtype, and strain levels.
• Demonstrated effective signal amplification from viral components using NIR-SERS.
• Overcame plasmonic mismatch issues common in conventional SERS techniques.

Conclusions:

• The AI-empowered NIR-SERS platform shows significant promise for enhancing rapid virus detection and identification.
• This approach addresses critical challenges in global public health preparedness by improving outbreak response capabilities.
• The system offers a generalizable diagnostic solution for identifying novel viral strains and subtypes.