This proposal aims to advance drug repurposing for rare cancers by utilizing cutting-edge computational approaches. The long-term objectives are to discover effective treatments for rare cancers with known drug targets and accelerate their transition into clinical practice. The proposed research is highly health-related as it addresses the urgent need for novel treatments for rare cancers, which pose a significant burden on affected individuals and their families. The company will construct a large-scale, high-quality biomedical knowledge graph (KG) using state-of-the-art natural language processing technologies to extract information from PubMed abstracts and PMC full-text articles. The approach will include the integration of data from public databases and genomics datasets, the development of an interpretable probabilistic inference algorithm, and the identification of combination therapies by minimizing potential adverse effects. The promising drug candidates identified will be experimentally tested using cancer cell line models in collaboration with business partners. The expected outcomes include a comprehensive KG tailored for drug repurposing, a robust and interpretable inference algorithm, a list of promising drug candidates for rare cancers, and experimental validation of selected candidates. The proposed research has the potential for significant technological innovation and may pave the way for breakthroughs in rare cancer treatments.