Leveraging knowledge graph embeddings for novel insights in colorectal cancer

Colorectal cancer is one of the leading causes of cancer death, and studies require advanced analysis methods to understand it more. This article proposes an automated end-to-end pipeline designed to systematically evaluate several established Knowledge Graph Embedding (KGE) models, such as TransE, RESCAL, ComplEx, and ConvE, to extend colorectal cancer research. Unlike most previous studies on isolated models, our approach examines their combined performance in biomedical applications in a comprehensive manner. In addition, we will introduce an enhanced triple classification framework, mentioning specifically the Threshold Calculation for Classification technique.Our pipeline performs efficiently on a medium-sized knowledge graph (28k triples), though scalability to larger-scale datasets remains to be tested in future work. The results indicate that ComplEx achieved the highest performance metrics for biomedical knowledge discovery among the models tested.