2026-10-06T09:48:50Z https://api.figshare.com/v2/oai

oai:figshare.com:article/34073650 2026-10-05T17:50:23Z category_7 category_19 category_734 category_931 category_811 portal_5 item_type_3 month_year_10_2026

<p>Full ranked lists of predicted synergistic (S4) and antagonistic (S5) drug pairs generated by the final HALO model for combinations absent from the training set.</p> Hannie Yousefabadi (25317274) Mahya Mehrmohamadi (4182847) Medicine Pharmacology Biological Sciences not elsewhere classified Information Systems not elsewhere classified Mathematical Sciences not elsewhere classified using strictly nested two independent datasets roc – auc realistic performance limits random data splits predicting antibacterial relationships performance dropped notably level chemical checker internal feature selection appear across folds 10 bacterial strains experimentally tested combinations antibacterial interaction learning scalable feature representation inflated performance metrics encoded using multi clinical bioactivity domains allow drug pairs rigorous evaluation framework halo </ b rigorous evaluation unseen combinations prediction framework learning approaches halo transferred validation schemes truly leakage strain variability results demonstrate permissive cross limited scale interpretable basis existing machine drug pair driven prediction current models curated dataset conservative conditions average precision 78 ). <p>Full ranked lists of predicted synergistic (S4) and antagonistic (S5) drug pairs generated by the final HALO model for combinations absent from the training set.</p> 2026-10-05T17:50:14Z Dataset Dataset 10.1371/journal.pcbi.1014849.s006 https://figshare.com/articles/dataset/_p_Full_ranked_lists_of_predicted_synergistic_S4_and_antagonistic_S5_drug_pairs_generated_by_the_final_HALO_model_for_combinations_absent_from_the_training_set_p_/34073650 CC BY 4.0