Computational framework identifies synergistic compound pairs in herbal medicines

Herbal medicines have long been used in disease treatment, often through formulas that combine multiple herbs to enhance therapeutic effects and reduce side effects. In recent years, network-based approaches have helped researchers map relationships among compounds, targets, and disease modules, providing new tools for drug discovery and drug combination prediction. However, herbal formulas contain hundreds or even thousands of ingredients, making it impractical to test all possible combinations experimentally. Moreover, the same formula may act differently across diseases, leaving the molecular mechanisms of herbal synergy insufficiently understood.

A study (DOI: 10.48130/targetome-0026-0013) published in Targetome on 26 March 2026 by Yinyin Wang's team, China Pharmaceutical University, reports a new framework that integrates network proximity and community analysis to identify disease-relevant synergistic ingredient pairs in herbal prescriptions.

To build the model, the researchers first selected five clinically used CVA-related herbal prescriptions: Suhuang Zhike Capsule, Huanglong Zhike Granule, Suzi Jiangqi Decoction, Shegan Mahuang Decoction, and Jiawei Dingchuan Decoction. Ingredient information was collected from seven traditional Chinese medicine databases, standardized using chemical structure identifiers, and filtered by oral bioavailability and drug-likeness. Ingredient targets were obtained from the STITCH database, while CVA-related genes were identified through transcriptomic analysis. The team then mapped herb-ingredient-target-disease relationships onto protein-protein interaction networks and calculated four types of interaction scores: herb-disease, ingredient-disease, herb-herb, and ingredient-ingredient proximity. Next, HerbSyner_Finder used the Louvain community algorithm to detect network modules in which ingredients or herbs clustered closely with the CVA disease node.

This approach generated a multidimensional "combination landscape" that highlighted potential synergistic components rather than isolated active compounds. Across the five prescriptions, the model identified kaempferol and quercetin as a broadly shared synergistic pair, while berberine and luteolin emerged as another strong candidate pair in the Shegan Mahuang Decoction-related network. The researchers then validated these predictions experimentally. In LPS-stimulated RAW264.7 macrophages, berberine and luteolin showed no obvious cytotoxicity at selected concentrations and jointly inhibited nitric oxide production more strongly than either compound alone. Their average synergy scores exceeded the threshold for synergy across multiple models, including ZIP, Loewe, HSA, and Bliss. In airway smooth muscle cells, the combination also inhibited LPS-induced abnormal proliferation, supporting its relevance to airway hyperresponsiveness.

In a CVA rat model, berberine-luteolin co-treatment reduced cough frequency, prolonged cough latency, alleviated inflammatory cell infiltration, reduced mucus secretion and fibrosis, and downregulated inflammatory markers including IL-1β, IL-4, IL-5, IL-6, TNF-α, MUC5AC, and COX-2. Mechanistic analysis further indicated that the pair suppressed NLRP3/NF-κB signaling, including reduced NLRP3 expression and decreased phosphorylation of p65 and IκBα.

Overall, this study presents HerbSyner_Finder as a practical computational framework for uncovering synergistic combinations in herbal medicines. By linking database mining, network biology, community detection, and experimental validation, the work provides a scalable route to explain the mechanisms of traditional formulas and identify promising natural compound pairs for future drug development.