KGWAS: novel genetics discovery enabled by massive functional genomics knowledge graph скачать в хорошем качестве

KGWAS: novel genetics discovery enabled by massive functional genomics knowledge graph

KGWAS is a novel geometric deep learning method that leverages a massive functional knowledge graph across variants and genes to significantly improve detection power in small-cohort GWASs. Thank you, Kexin Huang and Martin Jinye Zhang, for joining me on this one. Preprint: https://www.medrxiv.org/content/10.11... GitHub Page: https://github.com/snap-stanford/KGWAS Official Website: https://kgwas.stanford.edu/ 📖 Video Chapters: 0:00 - Introduction to Rare Disease Genetics Problem 0:54 - Welcome & Paper Overview 1:22 - What is KGWAS? The Elevator Pitch 2:41 - Why Finding More GWAS Hits Matters 4:46 - Origin Story: Collaboration with GSK 6:09 - Knowledge Graph Architecture Overview 7:36 - Building the Variant-Gene-Pathway Network 9:55 - Handling Linkage Disequilibrium (LD) Challenges 13:55 - Training the Large-Scale Graph Neural Network 14:16 - Model Training Requirements and Scalability 15:22 - Open Source Availability and Accessibility 16:05 - Scaling to Whole Genome Analysis 17:12 - Figure 2: Validation Through Simulations 18:24 - UK Biobank Downsampling Experiments 20:04 - Precision-Recall Performance Results 21:00 - Figure 3: Real-World Disease Applications 21:38 - Ulcerative Colitis Case Study Example 22:48 - Experimental Validation of Ulcerative Colitis Discovery 24:03 - Myasthenia Gravis Case Study Analysis 26:10 - Knowledge Graph Component Analysis and Ablation Studies 27:54 - Tissue-Specific vs Context-Agnostic Approaches 32:03 - Figure 4: Network Interpretation and Attention Mechanisms 36:30 - Alzheimer's Disease Network Visualization 40:01 - Drug Development Applications and Implications 41:15 - Post-GWAS Era Applications and End-to-End Solutions 43:01 - Figure 5: Compatibility with Existing GWAS Tools 48:16 - Population Diversity and Cross-Ancestry Applications 49:41 - Future Directions and Technical Improvements 52:16 - Input/Output Requirements and Computational Resources 54:28 - Website Demo and Interactive Features 56:59 - PhD Student Advice and Research Philosophy 59:36 - Closing Remarks