EPGFRE — Pan‑Genome Machine Learning for Fluoroquinolone Resistance in E. coli

EPGFRE (E. coli Pan‑Genome Fluoroquinolone Resistance Explorer) is a random forest model and accompanying dataset for predicting fluoroquinolone resistance from pan‑genome presence/absence profiles. This Zenodo record contains:- The trained Random Forest model (model.joblib)- The feature list (11,208 PGFam IDs) in training order (gene_features_list.pkl)- The binary presence/absence matrix (master_table.csv)- The resistance labels (labels.csv)- A sample genome matrix for testing (sample_genome_full.csv)- All validation result tables (cross‑validation, LODO, temporal, multi‑drug, permutation, etc.) Key results:- 5‑fold cross‑validated AUC: 0.914 ± 0.014- Permutation test: p = 0.001 (1000 shuffles)- 2,715 E. coli genomes, 11,208 gene families- Generalises to 38 additional drug classes (median AUC 0.91) The interactive web app is available at:https://huggingface.co/spaces/supzammy/EPGFRE Source code and analysis notebook:https://github.comzupzammy/EPGFRE Antibiogram source data (used for labels):https://doi.org/10.5281/zenodo.15809334