Bioactivity-driven prediction of antibacterial synergy using machine learning models
- 1Department of Molecular Biotechnology, School of Biotechnology, College of Science, University of Tehran, Tehran, Iran
Abstract
Motivation: Combination therapies have demonstrated improved efficacy across diverse clinical settings. Predicting antibacterial drug synergy remains difficult due to strain variability and the limited scale of experimentally tested combinations. Existing machine-learning approaches for synergy prediction often rely on permissive cross-validation schemes that allow drug pairs to re-appear across folds, leading to inflated performance metrics. A rigorous evaluation framework and scalable feature representation are needed for robust generalization. Results: We assembled a curated dataset of 3,160 drug–pair–strain interactions covering 97 antibacterial compounds and 10 bacterial strains. We then developed HALO (Held-out Antibacterial interaction Learning from latent bioactivity Observations), a synergy-prediction framework in which each drug pair is encoded using multi-level Chemical Checker (CC) similarity features spanning chemical, targets, networks, cellular, and clinical bioactivity domains. Using strictly nested, drug-pair heldout cross validation with fold-internal feature selection, HALO achieved consistent generalization to unseen combinations (ROC–AUC = 0.78). Performance dropped notably under increasingly stringent heldout schemes and inflated under random data splits, underscoring the importance of rigorous evaluation for this task. Despite these conservative conditions, HALO transferred to two independent datasets, achieving ROC–AUC = 0.90 and 0.71 and average precision = 0.76 and 0.94 for distinguishing synergy from antagonism. Together, these results demonstrate that multi-level bioactivity signatures provide a scalable, interpretable basis for predicting antibacterial relationships, while clarifying the realistic performance limits of current models under truly leakage-free evaluation. Author summary: Bacterial infections are increasingly hard to treat as resistance to antibiotics spreads faster than new drugs are developed. One strategy to fight back is combining two drugs together, since some pairs work better together than either does alone (synergy), while others can interfere with each other (antagonism). Predicting which drug pairs will be synergistic before running costly lab experiments could speed up the search for effective combinations. However, existing computational models are often evaluated in ways that let the same drug pair appear in both training and testing data, making their reported performance overly optimistic and unreliable in practice. We built a machine-learning framework called HALO that predicts antibacterial drug synergy using multiple layers of biological and chemical similarity information, while strictly preventing any drug pair from being seen during both training and evaluation. Under this stricter, more realistic testing setup, HALO still generalized well to drug pairs it had never encountered, and its predictions held up on two independent external datasets. Our results show that combining diverse bioactivity signatures with rigorous evaluation gives a more trustworthy — if more conservative — picture of how well these models can prioritize new antibacterial drug combinations for experimental testing.
Related
- GPCR Ligand Bioactivity Prediction with Physics-Informed Dual-State Query Learning
- Comparative phytochemical profiling and bioactivity evaluation of three Indonesian pomegranate (Punica granatum) peel varieties: Antioxidant, antibacterial, and in vitro antiproliferative activities
- Regional climate variability and rice yields prediction in Bangladesh using machine learning and deep learning models
- Integrating Multisource Target Evidence with Machine-Learning Potency Prediction for Alzheimer's Disease Drug Discovery
- An Explainable Hybrid Machine Learning Approach Based on MPSO-XGBoost for Pest Density Prediction in Almond Orchards
BibTeX
@article{yousefabadi2026bioactivitydriven,
title = {{Bioactivity-driven prediction of antibacterial synergy using machine learning models}},
author = {Hannie Yousefabadi and Mahya Mehrmohamadi},
journal = {PLOS Computational Biology},
year = {2026},
month = oct,
doi = {10.1371/journal.pcbi.1014849},
url = {https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014849},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Bioactivity-driven prediction of antibacterial synergy using machine learning models
AU - Hannie Yousefabadi
AU - Mahya Mehrmohamadi
PY - 2026
DA - 2026/10/05
JO - PLOS Computational Biology
DO - 10.1371/journal.pcbi.1014849
UR - https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014849
AB - Motivation: Combination therapies have demonstrated improved efficacy across diverse clinical settings. Predicting antibacterial drug synergy remains difficult due to strain variability and the limited scale of experimentally tested combinations. Existing machine-learning approaches for synergy prediction often rely on permissive cross-validation schemes that allow drug pairs to re-appear across folds, leading to inflated performance metrics. A rigorous evaluation framework and scalable feature representation are needed for robust generalization. Results: We assembled a curated dataset of 3,160 drug–pair–strain interactions covering 97 antibacterial compounds and 10 bacterial strains. We then developed HALO (Held-out Antibacterial interaction Learning from latent bioactivity Observations), a synergy-prediction framework in which each drug pair is encoded using multi-level Chemical Checker (CC) similarity features spanning chemical, targets, networks, cellular, and clinical bioactivity domains. Using strictly nested, drug-pair heldout cross validation with fold-internal feature selection, HALO achieved consistent generalization to unseen combinations (ROC–AUC = 0.78). Performance dropped notably under increasingly stringent heldout schemes and inflated under random data splits, underscoring the importance of rigorous evaluation for this task. Despite these conservative conditions, HALO transferred to two independent datasets, achieving ROC–AUC = 0.90 and 0.71 and average precision = 0.76 and 0.94 for distinguishing synergy from antagonism. Together, these results demonstrate that multi-level bioactivity signatures provide a scalable, interpretable basis for predicting antibacterial relationships, while clarifying the realistic performance limits of current models under truly leakage-free evaluation. Author summary: Bacterial infections are increasingly hard to treat as resistance to antibiotics spreads faster than new drugs are developed. One strategy to fight back is combining two drugs together, since some pairs work better together than either does alone (synergy), while others can interfere with each other (antagonism). Predicting which drug pairs will be synergistic before running costly lab experiments could speed up the search for effective combinations. However, existing computational models are often evaluated in ways that let the same drug pair appear in both training and testing data, making their reported performance overly optimistic and unreliable in practice. We built a machine-learning framework called HALO that predicts antibacterial drug synergy using multiple layers of biological and chemical similarity information, while strictly preventing any drug pair from being seen during both training and evaluation. Under this stricter, more realistic testing setup, HALO still generalized well to drug pairs it had never encountered, and its predictions held up on two independent external datasets. Our results show that combining diverse bioactivity signatures with rigorous evaluation gives a more trustworthy — if more conservative — picture of how well these models can prioritize new antibacterial drug combinations for experimental testing.
SN - 1553-7358
ER -