Bringing real-world data together

During patient diagnosis, treatment and clinical trials, data are captured using specialised tools, including electronic health record (EHR) platforms, clinical trial platforms and specialist scientific tools such as LIMS and ELNs. Data unification aims to de-silo data captured in the varied toolset while addressing data access issues. Increasingly, institutions and hospital systems are developing data aggregation solutions to bring data together. For example, multi-omic analysis platforms are being adopted to aggregate complex multi-omic data and allow for its interpretation in the context of the EHR. Unified views of this data in the correct context are crucial for the development of predictive models to inform treatment choices, the optimisation of trial design and enrolment, and for biomarker discovery through Genome-Wide Association Studies (GWAS).

To serve the use-cases, data not only needs to be brought together but harmonised. For example, if a researcher were looking for EHR and clinical trials referencing Breast Cancer, despite having access to all the required data, a string search for Breast Cancer would fail to return any data referencing Neoplasm of the breast or malignant mammary tumour, returning incomplete result sets.

The Center for Precision Medicine at City of Hope, one of the largest cancer research and treatment organisations in the United States, has created an enterprise-wide platform and precision medicine program to unlock the clinical value and research potential of complex and unique datasets by combining patient data with comprehensive genomic profiling and proprietary analytics. POSEIDON (Precision Oncology Software Environment Interoperable Data Ontologies Network) is a secure, cloud-based Oncology Analytics and Insights platform developed on DNAnexus®, a multi-omics and clinical analysis platform that enables exploration, analysis, visualisation, and collaboration on City of Hope’s patient clinico-genomic data along with public data sources.

POSEIDON has been leveraged by multiple research groups across oncology disciplines and institutions, driving insights that lead to individual patient care improvements. POSEIDON empowers scientists and clinicians to leverage omics data in a single, secure, cloud environment.

Data is aligned to these ontologies using TERMite, SciBite’s named entity recognition tool, which leverages all of these synonyms during entity recognition. Once entities are identified, they are assigned an ID that comes from the underlying ontology and enables the data to be aligned to the broader scientific community. In practice, instead of breast cancer, which could also be cancer of the breast, malignant mammary tumour or breast neoplasm, or a whole host of other synonyms, it is assigned the ID10006187 from the MedDRA ontology, and these additional synonyms are all associated with this ID. If users search or perform an analysis using any of these synonyms, they are associated with ID10006187 and results are returned, independent of which synonym was used to originally capture or search the data, known as semantic interoperability.

Another key consideration to success and adoption is long-term flexibility. Ontologies need to be dynamic and evolve alongside the field to continue supporting the data. The ability for research and treatment organisations like City of Hope, which includes more than 35 clinical locations in Southern California and clinical facilities in Arizona, Illinois and Georgia through its recent acquisition of Cancer Treatment Centers of America (CTCA), or for hospital networks to update public ontologies, such as the addition of a new drug name included in a clinical trial is paramount. CENtree, SciBite’s enterprise ontology management tool, is used by City of Hope to centrally manage, edit, and serve ontologies to POSEIDON. In CENtree, one may add additional entities or synonyms and push these to downstream systems, such as TERMite, to enable the periodic re-alignment of data to the updated standard, rather than extensive manual reprocessing of the data.