DeepDrug2: A Germline-focused Graph Neural Network Framework for Alzheimer’s Drug Repurposing Validated by Electronic Health Records

The preprint operates within the dominant paradigm of network-based drug repurposing for Alzheimer's disease, utilizing graph-based representations of biological interactions and applying machine learning for candidate prioritization.

The study refines and extends the existing DeepDrug framework, introducing methodological improvements within the established paradigm. It shifts focus from somatic to germline mutations, incorporates updated genetic data, and validates findings using real-world clinical data. These enhancements represent incremental progress within the current paradigm, rather than a challenge to fundamental assumptions or a shift to a new paradigm. The work adheres to established methodologies and evaluation metrics for GNN-based drug repurposing, further solidifying its classification as normal science. While introducing novel elements like updated biomedical graph construction and real-world clinical validation, these additions build upon and refine existing methods within the accepted paradigm, rather than proposing a radical alternative.