Leveraging Knowledge Graph and Imaging to Predict Gene–Disease Links in Cardiovascular Diseases
summary
Built a multimodal knowledge graph from UK Biobank plus 18 external databases that integrates cardiac MRI features as nodes with feature values stored on patient–CMR relationships, extracted CMR traits via a CNN, embedded the graph with a directed variational graph auto‑encoder and tested SVM, RF and ANN classifiers, with DVGAE and CMR nodes central to the pipeline. The KG contains about 33,000 nodes and 1.1 million relationships, PageRank gives CMR an average score of 51.1, and the SVM yields AUCs of 0.80 (HF), 0.78 (AF), 0.83 (MI) while top‑10 predicted genes show pathway enrichment (8/10 HF → 66 pathways, 7/10 AF → 14, 10/10 MI → 42).
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:06Background on gene-disease links
- 2:12Data sources and extraction
- 3:39Knowledge graph construction
- 5:06Embedding algorithm details
- 6:10Machine learning classification
- 7:15Enrichment analysis results
- 9:05Knowledge graph scale and accuracy
- 10:07Predicted genes per disease
- 11:33Feature importance analysis
- 12:37Impact of imaging data
- 13:20Summary and conclusions
- 14:23Q&A session
related talks
28mRole of Knowledge Graphs in Mental Health Diagnosis and Cure
33mNetwork Data Integration Enables Precision Medicine
39mIdentifying Disease Mechanisms Using Gene Networks as Priors
20mDiagnostic Medical application of ML
32mGraph-based ML for Automated Health Care Services
49mBuilding AI That Learns and Adapts: A Case Study in MRI Diagnostics
05mAugmenting Visual Information in Knowledge Graphs for Recommendations
27mAutomated Discovery of Cancer Types from Genes