BioSensGraph

This dataset contains a heterogeneous biomolecular knowledge graph designed for interaction prediction tasks. The graph integrates multiple types of biological entities, including proteins, peptides, nucleic acids, and small molecules, along with experimentally derived and similarity-based relationships. The dataset is distributed as CSV files organized into node and edge tables. The node tables describe molecular entities with associated metadata, while the edge tables define relationships between entities.The dataset includes six node files (AA.csv, DNA.csv, RNA.csv, NucleicMixed.csv, NucleicAmbigous.csv, SmallMolecule.csv) and two edge files (interacts_with.csv, has_similarity.csv). Nodes and edges are linked via a shared identifier (id), enabling reconstruction of the full property graph. The dataset can be used for link prediction, graph representation learning, and biomolecular interaction modeling.