RNA-KG views

Here, we provide RNA-KG views generated by our team. Their specifications are reported in a README.txt file. These views are thought to be used in combination with graph-oriented ML techniques for edge and node type labeling and heterogeneous/homogeneous link prediction tasks. For each view, the corresponding meta-graph (in pdf and xlsx formats) is provided along with two pickle dictionaries files for nodes and edges to facilitate the import in the GRAPE environment.

• View test: This view of test does not have 'semantic' value and should NOT be used as reliable or as a reference for predictions. The goal is to assign a confidence value to miRNA-disease and gene-miRNA triples.
• View miRNAdisease: It includes triples involving miRNAs, genes, and diseases. The goal is to assign a confidence value to miRNA-disease triples.
• View miRNAdiseaseSO: It includes triples involving miRNAs, genes, and diseases. Moreover, Sequence Ontology edges are integrated in the view. The goal is to assign a confidence value to the miRNA-disease triples and compare results with the miRNAdisease view.
• View RNAdiseaseGO: This complex view is designed for a heterogeneous link prediction task on edges of type miRNA-GO, lncRNA-GO, protein-GO, circRNA-GO, and lncRNA-disease.
• View piRNAdiseaseGO: This view is designed for a (directional) link prediction task on edges of type piRNA←→disease/phenotype.
• View lncRNAtRF: This view is designed for a link prediction task on edges of type lncRNA-anatomy, tRF-anatomy, lncRNA-cell, tRF-anatomy.
• View properties: It includes triples involving miRNAs, genes, histone modification, epigenetic changes, sequence features, phenotypes, and diseases. This view is enhanced with properties on nodes i.e. description and sequence, that can be exploited for improving link prediction methods. The goal is to assign a confidence value to miRNA-disease and miRNA-gene triples.