Direct detection and differentiation of the Vibrio harveyi clade using MALDI-TOF MS integrated with artificial intelligence for effective outbreak management

Abstract

Accurate identification of Vibrio harveyi clade species is critical for seafood safety and the control of aquaculture diseases. However, existing methods demonstrate limited classification performance. This study presents an artificial intelligence-assisted matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI–TOF MS) approach using on-plate protein extraction for direct colony analysis. Among six evaluated algorithms, the random forest model achieved perfect accuracy, significantly outperforming commercial databases, which showed only 79.57 % accuracy. Principal component analysis revealed improved species separation when using the top 10 % of features selected by the model. Receiver operating characteristic and precision-recall curves further confirmed the model's robustness and generalizability. Key mass peaks at 7194.7, 3609.8, and 9179.6 m/z were identified as major discriminatory features. This method offers a rapid, database-independent solution for classifying the V. harveyi clade species, enhancing MALDI–TOF MS utility in food safety and supporting its use in routine quality control.