LibKG

LibKG is a lineage-informed biological knowledge graph built for LINGO, a foundation model for single-cell lineage inference. It integrates genes, cell types, and pathways into a structured graph with eight biologically relevant relation types, including regulatory interactions, co-expression, differentiation, marker-gene links, protein interactions, and pathway associations. The current release provides species-specific LibKG resources for mouse, human, and zebrafish, including node tables, edge tables, and metadata used in the LINGO pretraining pipeline. LibKG is intended for lineage-aware pretraining, downstream lineage inference, and graph-based single-cell representation learning.