Raman spectroscopy is a nondestructive, label-free, highly specific approach that provides the chemical information on materials. Thus, it is suitable to be used as an effective analytical tool to characterize biological samples. Here we introduce a novel method that uses artificial intelligence to analyze biological Raman spectra and identify the microbes at a single-cell level. The combination of a framework of convolutional neural network (ConvNet) and Raman spectroscopy allows the extraction of the Raman spectral features of a single microbial cell and then categorizes cells according to their spectral features. As the proof of concept, we measured Raman spectra of 14 microbial species at a single-cell level and constructed an optimal ConvNet model using the Raman data. The average accuracy of classification by ConvNet is 95.64 ± 5.46%. Meanwhile, we introduced an occlusion-based Raman spectra feature extraction to visualize the weights of Raman features for distinguishing different species.

The need for efficient and accurate identification of pathogens in seafood and the environment has become increasingly urgent, given the current global pandemic. Traditional methods are not only time consuming but also lead to sample wastage. Here, we have proposed two new methods that involve Raman spectroscopy combined with a long short-term memory (LSTM) neural network and compared them with a method using a normal convolutional neural network (CNN). We used eight strains isolated from the marine organism Urechis unicinctus, including four kinds of pathogens. After the models were configured and trained, the LSTM methods that we proposed achieved average isolation-level accuracies exceeding 94%, not only meeting the requirement for identification but also indicating that the proposed methods were faster and more accurate than the normal CNN models. Finally, through a computational approach, we designed a loss function to explore the mechanism reflected by the Raman data, finding the Raman segments that most likely exhibited the characteristics of nucleic acids. These novel experimental results provide insights for developing additional deep learning methods to accurately analyze complex Raman data.