GMQL

GMQL enables advanced querying and integrative analysis of Next Generation Sequencing (NGS)-derived genomic, transcriptomic, and epigenomic region data by linking genomic features to experimental, biological, and clinical metadata for large-scale tertiary analyses.

Key Features:

• Declarative query language: A high-level language that abstracts genomic region data together with associated experimental, biological, and clinical metadata.
• Genomic Data Model (GDM): A unified framework that links genomic feature data with relevant metadata and describes semantically heterogeneous (epi)genomic datasets.
• File format support: Native interoperability with SAM, VCF, NARROWPEAK, and BED formats.
• Interoperability: Integration of diverse genomic, transcriptomic, and epigenomic datasets for cross-dataset queries.
• Scalability and execution engine: Implementation on the Hadoop framework and Apache Pig to enable scalable processing of large datasets.
• Intermediate representation: An intermediate representation that supports alternative implementations including Spark, Flink, and SciDB.
• Repository abstraction: A technology-independent abstraction capable of interfacing with local file systems, Hadoop File Systems, and databases.
• Downstream operation: Designed to operate downstream of raw data preprocessing pipelines and to query across thousands of heterogeneous datasets and samples.
• Tertiary data analysis: Support for tertiary analyses that investigate interactions and cooperations among different (epi)genomic regions and their products.
• Domain-specific computations: Enables development and execution of domain-specific computations and knowledge discovery workflows.

Scientific Applications:

• Integrative analysis of ENCODE data: Cross-dataset queries and integration of ENCODE genomic and epigenomic datasets.
• Roadmap Epigenomics integration: Comparative and integrative analyses of Roadmap Epigenomics datasets across tissues and marks.
• TCGA data analysis: Integration of TCGA genomic and transcriptomic data for tertiary-level studies of cancer genomics.
• Large-scale NGS studies: Big-data genomic analyses spanning thousands of heterogeneous samples for regulatory interaction and epigenomic profiling.

Methodology:

GMQL applies a high-level declarative query language over the Genomic Data Model (GDM), is implemented on Hadoop and Apache Pig, employs an intermediate representation enabling Spark, Flink, and SciDB backends, and uses a technology-independent repository abstraction for local file systems, Hadoop File Systems, and databases while operating downstream of raw data preprocessing pipelines.