SPOKE-related paper summaries
BioMedical Evidence Graph (BMEG) for Cancer
Exploring Integrative Analysis Using the BioMedical Evidence Graph. Struck A, Walsh B, Buchanan A, Lee JA, Spangler R, Stuart JM, Ellrott K. JCO Clin Cancer Inform. 2020 Feb;4:147-159. PMID: 32097025[back to paper list]
- cancer biology graph database and query engine
- “unique from other biologic data graphs in that sample-level molecular and clinical information is connected to reference knowledge bases”
- data on 15,000 patients, 52,000 samples, 6.8 million alleles, 640,000 drug-response experiments, and 50,000 literature-derived genotype-to-phenotype associations (> 41 million vertices and 57 million edges)
- query-based API client code available for Python, Javascript, and R
- online server bmeg.io and available on GitHub
BMEG Data Sources Include:
(see full list at BMEG website)
- The Cancer Genome Atlas (TCGA)
- Genotype-Tissue Expression (GTEx) [website]
- Cancer Cell Line Encyclopedia (CCLE) [website]
- Variant Interpretation for Cancer Consortion (VICC) Genotype to Phenotype (G2P) meta-knowledgebase of 12,856 aggregate clinical interpretations of cancer genomics covering 3,437 unique variants in 415 genes, 357 diseases, and 791 drugs [website]
- PharmacoDB with > 650K experiments involving 1600 cancer cell lines and 750 compounds [website]
- Drug Gene Interaction Database (DGIdb) [website]
- Cell Model Passports cancer cell model database
- Monarch Disease Ontology (MonDO) merged from multiple disease ontologies [website]
- MSigDB [website]
- Pathway Commons [website]
- Ensembl – IDs used to connect to GO, Pfam, ...
- PubMed, PubChem, ...
Figure: How example queries traverse the graph
(Project TCGA-BRCA, specific queries in
paper)
- Count Mutations per Gene in Breast Cancer
- Identify Pathways Containing Mutated Genes
- Determine the Number of Mutations in Each Pathway
- Find Publications Relevant to Phenotypic Consequences of Mutations
- Find Drugs Described in the Literature to Treat Phenotypes Linked to Mutations
- Find Drugs Tested in Breast Cancer Cell Lines
- Find the Sensitivity of Breast Cancer Cell Lines to a Drug
- Find Gene Expression Data Linked to Cell Lines
“To enable various analytical queries... we developed the Graph Integration Platform (GRIP)... [it] stores multiple forms of data and has the ability to hold thousands of data elements per vertex and per edge of the graph. This allows it to store sparse relationship data... as well as dense matrix-formatted data... The GRIP Query Language is a traversal-based graph-selection language inspired by Gremlin.”
Thoughts
- impressive amount of data in the graph
- significant learning curve for query language; not accessible to most researchers
- is it updated/updatable? maybe not
- are any of the data sources or ontologies of use to us?