Repurposing non-pharmacological interventions for Alzheimer’s disease through link prediction on biomedical literature

The complete workflow is depicted in Fig. 1. To investigate the association between NPIs and AD, we initially conducted preprocessing and integration of biomedical triples extracted from SemMedDB and SuppKG³³. Subsequently, we employed several graph representation models to derive the embedding information of ADInt. Ultimately, we selected the most effective model for generating hypotheses regarding and NPIs for AD and further evaluated them through the discovery patterns and RWD analysis.

Diagram illustrating the workflow of the methodology.

Materials

SemMedDB³⁰ is a repository of semantic triples extracted from PubMed abstracts and titles using the SemRep program³⁴. We obtained triples from the PREDICTION table of SemMedDB and the source sentences and text of triples from the SENTENCE and PREDICATION_AUX tables. This allowed us to supplement SuppKG with a broader range of…