Aggregating combinatorial biomedical graph ranking results for drug repurposi...
summary
Aggregates ranked drug–disease outputs from five orthogonal graph methods into an All-Against-All dataset covering 22,506 MONDO diseases and 52,882 chemicals, producing 1,157,416,190 ranked pairs. Methods include neural graph embeddings, rule mining, citation-count scoring, normalized co‑occurrence distance, and path‑based PageRank with EHR support. Implementation handles bulk queries and storage, exposes sparsity and API failure issues, and shows a top anti‑TNF hit for Castleman that aligns with proteomics and a concurrent trial. Plans focus on score fusion and prioritizing candidates for in‑vivo testing.
AI-written summary · about
chapters (10)
- 0:00Every Cure introduction
- 3:57Drug repurposing challenges
- 8:13Computational methods
- 12:04Translator results
- 18:07Every Cure timeline
- 19:55Clinical trial prediction
- 24:12Clinical relevance
- 27:45Metadata reproducibility
- 29:11Algorithm selection
- 30:56Q&A
related talks
05mPivot
30mBeyond the Silos Achieving Data Harmony With Graph Based ER
31mML methods to accelerate the drug discovery cycle
18mTranslating a Trillion Points of Data into New Insights into Disease
16mDocGraph and the Distribution of Open Datasets
10mA Story about Big Data in Healthcare
56mGraph Analytic Systems
32mData Mining in the EMR Gold Rush