Prediction of Antibiotic Susceptibility in E. coli Isolates Using Machine Learning.

Area of Science:

• Genomics
• Machine Learning
• Microbiology

Background:

• Antimicrobial resistance (AMR) is a critical global health issue, leading to millions of deaths annually and complicating treatments for bacterial infections like those caused by Escherichia coli.
• Current culture-based AMR detection methods are slow and lack precision, delaying crucial clinical decisions.
• Whole genome sequencing (WGS) presents a rapid and accurate alternative for AMR detection.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting the phenotypic antimicrobial susceptibility of Escherichia coli to ciprofloxacin using whole genome sequencing data.
• To explore the utility of combining different genomic feature types for enhanced prediction accuracy.

Main Methods:

• A novel dataset of 256 Escherichia coli genomes and associated antimicrobial susceptibility data was utilized.
• Genomic features were extracted using AMRFinderPlus (identifying known antimicrobial resistance genes) and k-mer frequency analysis (reference-free genomic patterns).
• Machine learning models, specifically Random Forest and XGBoost, were trained and evaluated using a five-fold cross-validation strategy.

Main Results:

• The study achieved over 90% accuracy in predicting Escherichia coli susceptibility to ciprofloxacin.
• Combining AMRFinderPlus results with k-mer frequency features significantly improved predictive performance.
• The XGBoost gradient boosting model demonstrated superior accuracy when utilizing this combined feature set.

Conclusions:

• Machine learning models, particularly XGBoost, can accurately predict antimicrobial susceptibility in Escherichia coli from whole genome sequencing data.
• Integrating known antimicrobial resistance gene markers with reference-free genomic features offers a powerful approach for AMR prediction.
• This WGS-based machine learning strategy holds promise for faster and more precise clinical decision-making in infectious disease management.