VITAGRAPH: building a knowledge graph for biologically relevant learning tasks

The complexity of human biology poses ongoing challenges, driving global interdisciplinary research. Artificial intelligence has become a powerful tool in computational biology, where graph data structures model entities like protein-protein interaction (PPI) networks and gene functional networks. These networks support crucial tasks in network medicine, including gene-disease association prediction, drug repurposing, and polypharmacy side-effect analysis. Reliable machine learning predictions require high-quality data. We present VITAGRAPH, a comprehensive multi-purpose biological knowledge graph built by integrating and refining multiple public datasets. Extending the Drug Repurposing Knowledge Graph, our pipeline: (a) resolves inconsistencies and redundancies, (b) consolidates information from leading public sources, and (c) enriches graph nodes with expressive features such as molecular fingerprints and gene ontologies. Incorporating biologically and chemically meaningful features enhances machine learning models' ability to learn accurate, structured embedding spaces. The resulting resource offers a coherent, reliable platform to advance computational biology and precision medicine while enabling benchmarking of graph-based models and offering the opportunity to tackle tasks such as drug repurposing, PPI prediction, and side-effect prediction, among others.