In a study published in Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, researchers from the Changchun Institute of Optics, Fine Mechanics and Physics of the Chinese Academy of Sciences, along with researchers from Jiangnan University and Wenzhou Medical University, developed a rapid method to identify lactic acid bacteria directly on agar plates combining adaptive Raman spectroscopy with a novel deep learning model.

Lactic acid bacteria play vital roles in food production, pharmaceuticals, and agriculture. Their identification methods like genetic testing and mass spectrometry require extensive sample preparation, destroy colonies, and cannot perform in situ detection. Existing Raman spectroscopy methods, which analyze molecular vibrations using laser light, struggle with biological heterogeneity within bacterial colonies, reducing the accuracy for industrial applications.

In this study, researchers proposed a two-part solution involving an adaptive colony Raman acquisition method based on signal-to-noise ratio screening. The technique automatically identifies optimal measurement points within colonies by analyzing spectral quality, and minimizes errors caused by uneven biological structures that previously hamper the identification. 

Then, researchers paired this technique with the Raman Swin Transformer model, a deep learning model adapted from image-processing algorithms. This model can efficiently capture subtle spectral patterns through a specialized attention mechanism to analyze local and global features in Raman data.

Using a near-infrared laser, the system scanned bacterial colonies in situ through agar plates with each measurement taking half a second. Researchers analyzed the spectral data of fourteen lactic acid bacteria strains. The Raman Swin Transformer model achieved 98.2% accuracy in species classification. The model maintained over seventy percent accuracy in identifying new strains of previously learned species from different sources. Besides, it outperformed other models in robustness tests.

The developed non-destructive method enables real-time selection of bacterial colonies for industrial applications like dairy fermentation and probiotic development. Reducing the time for strain screening from days to minutes could streamline microbial resource banking and quality control in biotechnology.