Application of MALDI-TOF MS-based peptidome profiling for the identification of Bacillus cereus, Staphylococcus aureus, and Escherichia coli in single and mixed inoculum.

Abstract

The detection of mixed-species bacterial samples plays a vital role in ensuring food safety, yet research in this area remains notably limited. This study investigates the integration of MALDI-TOF MS-derived peptidome profiles with artificial intelligence (AI) to enable accurate identification of mixed bacterial species. The application of formic acid extraction, combined with a residual network (ResNet) significantly improved the prediction accuracy for mixed bacterial samples in multi-scale complex datasets, achieving 96.88 % accuracy in identifying strains from three species: Bacillus cereus, Staphylococcus aureus, and Escherichia coli. Furthermore, peptidome profiling annotation was performed using label-free proteomics based on liquid chromatography-tandem mass spectrometry (LC-MS/MS) and differential protein screening using a random forest-convolutional neural network (RF-CNN) further enhanced strain identification. This advanced strategy improves the detection of targeted strains in mixed samples, supporting MALDI-TOF MS applications in the food industry.