Knowledge Graph and Machine Learning Help the Research of Drugs Aimed at Neurological Diseases

Abstract

In this study, we present PATHOS (PATHologies of HOmo Sapiens), a semantically rich knowledge graph constructed by integrating diverse datasets spanning multiple biomedical entity types. PATHOS provides a comprehensive resource for representing and exploring the intricate relationships underlying human diseases. To leverage this resource, we developed LOGOS (Learning Optimized Graph-based representations of Object Semantics), a graph embedding model capable of generating predictions relevant to drug research. The PATHOS-LOGOS framework was validated through three neurological case studies: drug repurposing for Alzheimers disease, phenotype selection for Huntingtons disease, and protein target identification in multiple sclerosis. The results demonstrate the potential of this approach to advance therapeutic insights and inform biomedical research.