GenomicsDB

GenomicsDB provides scalable, tile-based sparse-array storage and high-performance query capabilities for genomic variant data to support large-scale genomic analysis and precision medicine.

Key Features:

• High-Performance Data Storage: Implemented in C++ and leveraging a fork of htslib, GenomicsDB uses a tile-based sparse array storage system optimized for handling variant data and supports importing, querying, and transformation of variant records.
• Distributed Platform Support: Operates with a shared-nothing architecture and Single Process Multiple Data (SPMD) approach to read and write variant records in parallel across multiple processes, enabling scalability as sample counts increase.
• Java/Spark Integration: Integrated with Java and Apache Spark to enhance performance in distributed computing environments.
• ODA Framework Integration: Extended for integration within the ODA distributed analytics framework for clinical and genomic data integration.
• Scalability and Efficiency: Exhibits worst-case linear scaling in storage size, import time, and query execution as the number of samples increases.
• Precision Medicine Support: Enables analysis of biomedical datasets to identify personalized diagnosis and treatment options based on shared clinical and genetic features.
• HIPAA Compliance: Deployed in a HIPAA-compliant manner for handling sensitive medical data.

Scientific Applications:

• Precision Medicine: Integration and analysis of genomic and clinical data to support personalized diagnosis and treatment decisions.
• Large-Scale Variant Analysis: Cohort-scale storage, import, and query of variant datasets for population- and clinical-scale studies.
• Distributed Clinical-Genomic Analytics: Support for distributed analytics on combined clinical and genomic datasets within the ODA framework.

Methodology:

Implements a C++ tile-based sparse array storage using a fork of htslib; supports importing, querying, and transformation of variant records; runs in a shared-nothing architecture with SPMD parallel read/write; integrates with Java/Spark and the ODA framework; and demonstrates worst-case linear scaling for storage size, import time, and query execution.