An Ontology‑Guided Drug–Herb–Food Interaction Checker with Mechanism‑Based Knowledge Graph Reasoning and Condition‑Aware Interpretation

<ns5:p>Background The concurrent use of prescription medicines with herbal products, dietary supplements, and foods is common, particularly among individuals with chronic diseases. Such real-world co-consumption generates interaction patterns beyond conventional drug–drug interactions. Existing interaction-checking systems remain largely drug-centric, rely on predefined interaction pairs, and provide limited mechanistic transparency and condition-aware interpretation. Consequently, they are poorly equipped to represent interactions influenced by health-related conditions such as age, renal impairment, pregnancy, and lifestyle. Methods We developed the Drug–Herb–Food Interaction Checker (DHFI-C), an ontology-guided knowledge graph platform for mechanism-based and condition-inclusive interaction assessment. Evidence was curated from open-access literature under PRISMA 2020 and transformed into a structured data model spanning drugs, herbs, foods, health-related conditions, and underlying diseases. Entities and interaction components were aligned with external biomedical ontologies where appropriate, whereas a DHFI mini-ontology captured underrepresented interaction concepts. The model was implemented as a graph-native representation paired with a deterministic inference engine that derives pharmacokinetic and pharmacodynamic interactions through shared mechanistic pathways. We evaluated the DHFI-C using a comprehensive, predefined use case. Results The knowledge graph integrated &gt;24,000 drug entities, 92 herb/food entities, 1,277 curated interaction records, and 5,294 mechanism-inferred interaction records, resulting in 6,571 total retrievable interaction candidates, along with mechanistic nodes for enzymes, transporters, and pharmacodynamic effects. The DHFI-C reports both curated and mechanism-inferred interactions with explicit provenance. In the use case, the system handled multi-domain interactions, produced condition-level interpretations, detected pharmacological effect duplication, decomposed combination products, and supported disease-driven drug suggestions. Outputs are available in consumer and expert presentation modes, with mechanistic explanations. Conclusions DHFI-C provides a transparent and extensible framework for drug–herb–food–condition interactions. By integrating multidomain data within a mechanism-oriented knowledge graph, it enables context-aware interpretation and generation of mechanism-inferred interaction candidates.</ns5:p>