BMEG

BMEG integrates heterogeneous cancer-related molecular, clinical, and drug-response datasets into a large biomedical evidence graph to enable integrative analyses in cancer biology.

Key Features:

• Graph database and query engine: Implements a heterogeneous graph database and query engine to represent and traverse complex biological relationships.
• Integrated resources: Aggregates data from over 20 large-scale resources.
• Supported data types: Represents RNA sequencing, genome-wide copy number variations, DNA methylation profiles, somatic mutations from whole-exome or whole-genome analyses, pathology estimates, drug response outcomes, and clinical and phenotypic measurements.
• Linkage to knowledge bases: Connects sample-level molecular and clinical information with reference knowledge bases, including pathway databases and literature-derived associations.
• Scale: Contains over 41 million vertices and 57 million edges.
• Dense feature vectors and complex relationships: Supports representation of dense feature vectors and complex, multi-entity relationships for analysis.
• Query-based API and clients: Exposes a query-based API with client libraries available for Python, JavaScript, and R to perform programmatic analyses.
• Cross-dataset analysis: Enables analyses that span multiple datasets within the integrated graph.

Scientific Applications:

• Mutation significance analysis: Assess mutation patterns and significance across integrated cohorts and datasets.
• Drug-response machine learning: Support drug-response modeling and machine-learning analyses using linked molecular and phenotypic data.
• Patient-level knowledge-base queries: Query patient-specific molecular and clinical profiles against reference knowledge bases and literature-derived associations.
• Pathway-level analysis: Analyze pathway associations and pathway-level alterations by linking molecular data to pathway databases.

Methodology:

Construction and storage as a heterogeneous graph database and query engine integrating multiple datasets, with analysis performed via a query-based API and client libraries in Python, JavaScript, and R.