Accurate virus identification with interpretable Raman signatures by machine learning

Abstract

Significance A large Raman dataset collected on a variety of viruses enables the training of machine learning (ML) models capable of highly accurate and sensitive virus identification. The trained ML models can then be integrated with a portable device to provide real-time virus detection and identification capability. We validate this conceptual framework by presenting highly accurate virus type and subtype identification results using a convolutional neural network to classify Raman spectra of viruses. The accurate and interpretable ML model developed for Raman virus identification presents promising potential in a real-time, label-free virus detection system that could be used in future outbreaks and pandemics.