A knowledge graph–driven literature-based discovery framework for drug repurposing in aplastic anaemia

Background Aplastic anaemia is a haematological disorder characterized by immune-mediated destruction of haematopoietic stem cells and bone marrow failure. Although immunosuppressive therapies are currently used, their lack of target specificity underscores the need for systematic approaches to identify precise therapeutic targets and explore drug repurposing opportunities. Methods A hybrid computational framework integrating literature-based discovery and database-driven validation was employed to construct a multi-layered biomedical knowledge graph. Semantic predications were extracted from SemMedDB and filtered using Unified Medical Language System semantic types to identify disease-, gene-, and drug-related entities. A tripartite disease–gene–drug network was constructed and subsequently extended into a heterogeneous knowledge graph by incorporating pathway-level interactions derived from enrichment analysis using clusterProfiler and KEGG. Drug–gene associations were further supported using curated resources such as DrugBank and DGIdb. Validation of gene–disease associations was performed using DisGeNET. Results The constructed knowledge graph revealed key genes, including CSF2, IL6, EPO, HLA-DRB1, and CD4, which are central to immune dysregulation and haematopoietic impairment. Pathway enrichment analysis identified critical pathways such as haematopoietic cell lineage, JAK–STAT signaling, cytokine–cytokine receptor interaction, and MAPK/ERK cascades, highlighting the interplay between immune signaling and bone marrow dysfunction. Integration of drug-related predications enabled the identification of both clinically used immunosuppressants and potential repurposing candidates, including mycophenolate mofetil, mizoribine, everolimus, and camostat mesylate. Mechanistic mapping of these drugs to key pathways suggests their potential to modulate aberrant immune responses and restore haematopoietic balance. Conclusion This study demonstrates the utility of a knowledge graph–based framework for integrating heterogeneous biomedical data to uncover multi-level interactions in aplastic anaemia. The approach enables systematic identification of biologically relevant genes, pathways, and drug candidates, supporting hypothesis-driven drug repurposing. While the findings are supported by literature and database validation, they remain exploratory and warrant further experimental and clinical investigation.