The abstract is read at the publisher; the summary is JClub's.
Ampicillin and trimethoprim-sulfamethoxazole resistance in Escherichia coli are discriminated by genomic-feature machine learning, reaching held-out accuracies of 93.98% and 93.59% with ROC-AUC up to 0.954.
- Why it matters: AMR hampers treatment decisions, and genomic predictors must be tested for both predictive accuracy and generalisability, including how well they detect resistant isolates.
- What they did: Using 446 Escherichia coli genomes, 252 binary genomic features, and 1,935 susceptibility observations across nine antibiotics, logistic regression, random forest, and XGBoost were validated by ROC-AUC and then tested on held-out subsets.
- The result: Ceftazidime showed 88.52% accuracy and 0.940 ROC-AUC but only 22.22% resistant-isolate recall, while removing overlapping observations reduced ROC-AUC from 0.912 to 0.848 (ampicillin) and 0.954 to 0.901 (trimethoprim-sulfamethoxazole).
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.