Genomics Research Capabilities
Genomics represents a prime example of “big data” in biomedical research. Each individual’s genome contains approximately 6 billion data points and harbors 4-5 million genetic variants compared to the human reference genome. Identifying which of these variants are causative in pediatric disease is a monumental task that requires sophisticated computational solutions.
Enterprise Genomics Applications
Under ODS’s leadership, Nationwide Children’s has developed a comprehensive, clinically validated suite of tools that manage the entire genomics lifecycle, from sample receipt and sequencing to final interpretation and reporting. This integrated ecosystem comprises three core components — Phoenix, Churchill and Varhouse — that are integral to patient care at Nationwide Children’s. Rigorously validated under CAP/CLIA and New York State CLEP standards, this system is used for all patients undergoing diagnostic testing and tumor molecular characterization in the Institute for Genomic Medicine. The impact of this platform is profound: by helping clinicians and researchers pinpoint the causes of cancers and rare diseases, it has provided critical answers for thousands of families, many of whom had endured long diagnostic odysseys. Moreover, it serves as a model for robust genomically informed care for other health systems and stands as a powerful example of how ODS is successfully translating data science innovation into tangible improvements in patient outcomes.
Enterprise Genomics Infrastructure
Phoenix is the operational backbone of our genomics pipeline. As the central hub for all genomic samples, it provides end-to-end tracking and management throughout the entire sequencing process. Phoenix serves as the primary orchestrator for our complex bioinformatics workflows, ensuring that each sample is processed efficiently, reproducibly and according to the highest quality standards. By automating and managing the flow of data from the sequencer to our analysis pipelines, Phoenix reduces manual effort and ensures the integrity of the data that fuels discovery.
Once orchestrated by Phoenix, raw sequencing data is processed by Churchill, a comprehensive suite of ultra-fast, deterministic and highly scalable bioinformatics pipelines. Churchill is the computational engine responsible for transforming raw data into meaningful genomic insights. Its capabilities include:
- Germline Variant Discovery: Identification of single nucleotide variants (SNVs), small insertions or deletions (indels), copy number variations (CNVs) and structural variants (SVs) from whole genome and exome data.
- Somatic Variant Detection: Specialized pipelines for tumor-normal sequencing analysis to identify acquired mutations critical in understanding and treating.
- Transcriptome Analysis: RNA-Seq pipelines for detecting gene fusions and performing expression profiling to provide a more complete molecular picture of a patient’s disease.
All processed data and discovered variants are loaded into Varhouse, a state-of-the-art, cloud-based genomic data warehouse. Built on modern big data technologies like Apache Spark and AWS infrastructure, Varhouse offers virtually limitless storage and parallel processing capabilities, serving as the cornerstone for both clinical diagnostics and translational research for tens of thousands of subjects.
The Varhouse platform provides clinicians and researchers with a powerful web-based interface to filter, prioritize, annotate and visualize genomic variants. It automates the reanalysis of existing data against the latest scientific literature and annotations, ensuring that interpretations remain current.
Built into Varhouse are advanced AI tools designed to aid in the most challenging aspect of genomics: variant interpretation. A key innovation is the Clinical Annotation of Variants with Language Representations (CAVaLRi), a sophisticated machine learning algorithm that integrates clinical characteristics from patient notes with genomic data. By using natural language processing to understand a patient’s unique clinical features, CAVaLRi can identify and prioritize the most likely diagnostic variants for review.
Research Genomics Infrastructure
To foster continuous innovation of Nationwide Children’s genomics capabilities, ODS is building a dedicated Research Genomics Data Lake. Part of ODS’s broader Data Lake strategic pillar, the Research Genomics Data Lake sits within a secure and highly scalable cloud computing environment built on Amazon Web Services (AWS) petabyte-scale services. Large-scale clinical and research-generated sequencing data is automatically ingested into the Data Lake environment, whose serverless architecture allows the data to be processed and analyzed rapidly and cost-efficiently while preserving privacy, maintaining robust system security, and allowing agile integration of new cloud services. AWS provides the flexibility to scale up to thousands of instances simultaneously, enabling on-demand execution of computational tasks that are massively parallel or have high compute requirements. This “Infrastructure-as-a-Service” model allows for the creation and customization of complex computing environments entirely through software. This proficiency has led to a partnership with AWS product teams, allowing ODS to influence the development of future AWS services.
In the Data Lake, genomic data from individual patients can be linked to diverse institutional data sources, including EHRs, imaging, physiologic monitoring, and longitudinal outcomes, to yield comprehensive patient records for use in genomics research. The Data Lake environment is equipped with gold-standard genomic data processing capabilities for variant calling, joint/cohort-based genotyping, and high-quality, near-real-time annotation. This pipeline relies on an ever-evolving suite of cutting-edge tools, such as the Dynamic Read Analysis for Genomics (DRAGEN)-Genome Analysis Toolkit (GATK), which merges Illumina's hardware-accelerated data analysis platform with the Broad Institute's established toolkit for variant discovery. Data querying and genotype-phenotype mapping are facilitated by a user-friendly, AI-powered interface based on the DREAM architecture. In addition to data storage, linkage, and analysis, the Data Lake environment serves as a testbed for continual evaluation of emerging genomic data processing tools and seamless integration of new tools into existing bioinformatic workflows.
With its deep reservoir of multidimensional data, state-of-the-art analytic capabilities, and secure design, the Research Genomics Data Lake is a strategic asset for Nationwide Children’s, asserting it as a trusted partner of choice for translating genomic discoveries into real-world impacts for patients. As an institutional resource, the Research Genomics Data Lake will position Nationwide Children’s and its research partners as leaders in identifying novel genetic mechanisms of disease, improving precision diagnosis and therapeutic development, and modernizing the practice of health outcomes research.