Machine learning-enhanced MALDI-TOF MS for real-time detection of antibiotic-resistant E. coli in food processing

• Published in: Food science & technology Vol. 224; p. 117860
• Main Authors: Victor Lin, Hong-Ting, Yang, Tien-Wei, Lu, Wen-Jung, Chiang, Hong-Jhen, Hsu, Pang-Hung
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 15.05.2025
• Subjects: algorithms; antibiotic resistance; Antimicrobial resistance; carbapenems; chloramphenicol; desorption; Escherichia coli; food production; Food safety; Machine learning; MALDI-TOF MS; mass spectrometry; monitoring; penicillins; prediction; public health; quinolones; Rapid detection; rapid methods; risk; streptomycin; tetracycline; MALDI-TOF MS; Food safety; Antimicrobial resistance; Rapid detection; Machine learning
• DOI: 10.1016/j.lwt.2025.117860
• ISSN: 0023-6438

Antibiotic-resistant Escherichia coli in food processing poses a significant risk to public health, necessitating rapid detection methods. This study developed an innovative approach combining matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) with machine learning for rapid detection of antibiotic-resistant E. coli in food processing environments. Analysis of 69 E. coli isolates from food processing facilities revealed high resistance rates, ranging from 0 % for carbapenems to 100 % for antibiotics like streptomycin and sulfamethoxazole-trimethoprim. These findings highlight serious food safety concerns and emphasize the need for rapid detection methods. Among machine learning models trained on MALDI-TOF MS data, the optimized random forest model demonstrated superior performance, achieving cross-validation accuracies within 67–97 % across different antibiotics. Validation using 28 food-sourced samples confirmed its high predictive accuracy for multiple antibiotic classes, including penicillin, chloramphenicol, sulfonamide, tetracycline, and aminoglycoside. This approach provides a rapid, accurate tool for antibiotic resistance detection, offering significant advantages for food safety monitoring in high-throughput processing environments. Future improvements should focus on enhancing (fluoro)quinolones prediction accuracy to enable comprehensive antimicrobial resistance surveillance in food production. •Developed MALDI-TOF MS machine learning approach for antibiotic-resistant E. coli detection.•Found high antimicrobial resistance rates in poultry facility bacterial isolates.•Optimized random forest classifier achieved best resistance prediction performance.•Validated 87–100 % detection accuracy on food-sourced E. coli isolates.•Created rapid detection method compatible with current food safety protocols.