A Novel Computational Model for Traditional Medicine Knowledge Formalisation and Drug Interaction Prediction from Nigerian Ethnopharmacological Data

Nigerian traditional medicine practitioners possess a rich body of plant-based pharmacological knowledge that is undocumented in machine-readable form, limiting its integration into evidence-based clinical practice and drug interaction safety research. This study develops a novel computational model for traditional medicine knowledge formalisation and drug interaction prediction from Nigerian ethnopharmacological data. Four original contributions are presented: a Nigerian Traditional Medicine Ontology (NiTMO) covering 1,200 medicinal plant species, 3,400 traditional use descriptions, and 850 active compound mappings drawn from ethnopharmacological surveys across six geopolitical zones; a natural language processing pipeline for automated extraction of plant-condition-treatment triples from Yoruba, Hausa, and Igbo traditional healer interview transcripts; a graph-based drug-herb interaction prediction model using a biomedical knowledge graph combining NiTMO with DrugBank and ChEMBL data, trained on known interaction data and evaluated on a curated Nigerian interaction test set; and a probabilistic safety risk scoring system for traditional-modern drug combination scenarios. The interaction prediction model achieved an AUC-ROC of 0.883 on the hold-out interaction test set. Three high-severity previously undocumented interactions were identified and validated through literature search and in vitro screening. NiTMO is publicly released for community extension. The study constitutes an original contribution to computational ethnopharmacology and provides infrastructure for integrating Nigerian traditional medicine into evidence-based drug safety monitoring.

Keywords: traditional medicine, drug interaction, knowledge graph, Nigerian ethnopharmacology, computational pharmacology