Knowledge graph integration of clustered medicinal plants, molecules, diseases, and targets

This study builds on the premise that phytochemicals, when co-existing across multiple plant species, tend to form structural and functional clusters. While individual compounds may exhibit distinct pharmacological properties in isolation, their behaviour within molecular clusters often diverges, potentially leading to emergent synergistic effects. Leveraging this insight, we systematically analysed 490 phytoconstituents derived from ten medicinal plants belonging to the Dasamoola group. Through chemoinformatic clustering using K-Means and dimensionality reduction via t-SNE, these molecules were organised into 49 structurally coherent clusters. Pharmacological relevance was assessed by mapping clusters to 87 ICD-11-classified disease conditions, thereby integrating clustering, ICD-11 mapping, and knowledge-graph visualization into a unified workflow that can serve as a template for analysing other complex polyherbal formulations. Heat map analyses revealed significant correlations between molecular clusters and disease phenotypes, indicating potential poly pharmacological mechanisms. To further elucidate these relationships, predicted molecular targets were integrated with disease ontologies using a Neo4j-based knowledge graph framework. This network-based approach enabled the visualization of molecule-target-disease associations, suggesting mechanistic insights that extend beyond conventional reductionist perspectives in an exploratory manner. Overall, our findings suggest potential molecular-target-disease associations that link structurally related phytochemicals to defined disease categories through shared biological targets. These associations indicate plausible network-level relationships and offer new avenues for, understanding the systems-level pharmacology of traditional medicinal formulations, generating testable hypothesis that warrant further experimental validation.