A novel framework has been proposed for integrating multi-modal data in Alzheimer's disease research using large language models and knowledge graphs.
The approach allows for population-level integration of various data types from independent cohorts without the need for matched patient IDs.
Statistical analysis identified significant features in each modality, which were connected as nodes in a knowledge graph for further analysis.
The framework revealed novel relationships and correlations between different data modalities, providing new insights into Alzheimer's disease pathology.