BioKMS-HAG: A hierarchically guided biomedical and space science knowledge fine-grained mining system.

The rapid growth of biomedical data and advancements in artificial intelligence have created unprecedented opportunities for scientific discovery. However, effectively leveraging these technologies requires integrating cutting-edge methods into accessible tools. In this study, we develop the Biomedical Knowledge Mining System (BioKMS) that integrates knowledge graph construction and advanced link prediction with the Hierarchy Guided Approach for Fine-Grained Biomedical Hypothesis Generation (HAG). BioKMS provides an interactive platform for exploring biomedical and space life science data, while HAG refines link prediction through a hierarchical tree-guided Graph Neural Network model, enabling fine-grained hypothesis generation with reduced computational overhead. By integrating HAG into BioKMS, we demonstrate significant improvements in predictive accuracy, particularly in distinguishing complex and confusing biomedical entities. The integration supports seamless knowledge discovery across disciplines, showcasing its potential to streamline hypothesis generation and accelerate research efficiency This work highlights the transformative potential of combining advanced AI methodologies in user-friendly systems, paving the way for more efficient biomedical and space science research.