AI Models Accurately Predict E. coli Antibiotic Resistance in Agricultural Environments, Enhancing Food Safety Monitoring

Escherichia coli (E. coli) is a widespread bacterium found in the intestines of both animals and humans. It is frequently used to detect fecal contamination in the environment and is known for its ability to rapidly develop antibiotic resistance. This characteristic makes E. coli a valuable tool for studying antimicrobial resistance, especially in agricultural settings where manure and wastewater are commonly reused. Traditional methods for analyzing antimicrobial resistance, such as disk diffusion assays, are often time-consuming and labor-intensive, limiting their practicality for large-scale monitoring. To address this, researchers are investigating alternative approaches that combine whole-genome sequencing (WGS) with predictive modeling using artificial intelligence (AI). Marco Christopher Lopez and Dr. Pierangeli Vital, from the University of the Philippines–Diliman College of Science's Natural Sciences Research Institute (UPD-CS NSRI), collaborated with Dr. Joseph Ryan Lansangan from the UPD School of Statistics. They evaluated several AI prediction models to determine the antimicrobial resistance of E. coli using genetic data and lab test results from the National Center for Biotechnology Information (NCBI) database. "Models were selected based on their ability to handle biological and imbalanced data," Dr. Vital explained. "We aimed to compare different learning strategies and identify the most suitable model for predicting antibiotic resistance." The AI models used included Random Forest (RF), which is adept at managing high-dimensional data; Support Vector Machine (SVM), which excels in classification tasks with complex decision boundaries; and two ensemble methods—Adaptive Boosting (AB) and Extreme Gradient Boosting (XGB)—which improve accuracy by focusing on difficult-to-classify samples. The models performed exceptionally well in predicting resistance to streptomycin and tetracycline, achieving high accuracy and effectively distinguishing resistant strains from susceptible ones. However, predicting resistance to ciprofloxacin proved challenging due to the scarcity of resistant samples (only 4% of the dataset). Despite this, AB and XGB showed consistent accuracy across various types of imbalanced data. "This approach holds significant promise for real-time monitoring of antimicrobial resistance in agriculture," Dr. Vital stated. "With DNA sequencing becoming increasingly rapid and cost-effective, our models can help detect resistant bacteria early, preventing outbreaks and enhancing food safety, agricultural practices, and public health initiatives." To further refine their predictions, the researchers recommend incorporating a broader range of sample types and data sources, such as metagenomic data, which includes DNA from all microbes present in a sample. This diversification could provide a more comprehensive understanding of how bacteria develop resistance. Dr. Vital also emphasized the importance of interdisciplinary collaboration, noting that microbiologists and statisticians working together produced a more impactful outcome. "Integrating biological concepts into statistical and predictive modeling can yield significant benefits for agricultural food safety and the wider community," she added. The research findings are detailed in the Malaysian Journal of Microbiology, with the paper titled "Prediction Models for Antimicrobial Resistance of Escherichia coli in an Agricultural Setting Around Metro Manila, Philippines." The study not only highlights the potential of AI in monitoring antimicrobial resistance but also underscores the importance of collaborative efforts in advancing this field. DOI: 10.21161/mjm.240650 Provided by University of the Philippines–Diliman