1. Clinical research data interoperbility Shared names meeting, Boston, 2009-04-29 Bosse Andersson (AstraZeneca R&D Lund) Kerstin Forsberg (AstraZeneca R&D Mölndal) Creative Commons License: allowed to share & remix, but must attribute & non-commercial

2. Main expectation Better interoperability, e.g. secondary use of data, also for new purposes and easier integration Better provenance, leading to more informed decisions (trust) Better utilization of computers, e.g. reasoning Better externalisation, e.g. utilize Linking Data principles for integrating internal and external data Clinical research data interoperability #2

3. RA Cancer Alzheimer’s Disease COPD Asthma Chronic Pain CVD Co-morbidity between diseases Clinical research data interoperability

4. Semantic Data Integration for Early Clinical Development Use Case Improve the capability to integrate and interpret heterogeneous data – Evaluate hypotheses about patient characteristics and other factors that can explain segmentation criteria – Data interpretation is a non-trivial process that requires overcoming: Semantic differences in the format, e.g. used identifiers Verify, validate and compare experimental results with other established data sets Efficient secondary usage of past experimental results and analysis conducted in later phases – Signal evaluation of adverse drug event reports evaluate if there is a casual relationship between the drug and the adverse event Clinical research data interoperability #4

5. Consolidated clinical data repository The CRL and CCDS are designed based on the assumption that diversity in clinical data is part of “doing research”. – Driver: Business value achieved by effective use of clinical data cross studies and over time – So, in CRL we will be able to specify the variances in what we observe on subjects in clinical studies, and the information about these observations. – CCDS will connect these specifications to the actual data. And thereby enable us to take informed decisions when we want to utilize data cross variances. Enforcement of standards to reduce diversity is a line organization decision. – Driver: Operational efficiency by rationalization of processes and tools for new studies. – So, CRL will make this task easier by making the preferred (standardized) variant of the specification available as first option when we will set up new studies and acquire new information. Existing Studies New Studies Clinical research data interoperability

6. Clinical Observation Concepts To store the clinical observation within the CRL data model we need to define some terminology What are we trying to measure? Systolic Blood pressure (carrier of topic ) Could it be measured in a different way and would that affect the result? YES Patient position (qualifier) Method/Tool/Equipment (qualifier) Location/Site -where you measure it (qualifier) For the clinical trial is there anything I need to know? YES When was it measured, date (context) Concepts Clinical research data interoperability

7. Linking Open Drug Data A significant challenge within life science and health care is a strong prevalence of terminology conflicts, synonyms and homonyms. Expressive querying or reasoning require deeper semantic integration. Clinical research data interoperability

8. Acknowledgement Ontotext – Vassil Momtchev LarKC consortium / LarKC is partially funded by EU FP7-215535 LODD team in HCLS IG #8 Clinical research data interoperability

9. Large Knowledge Collider in a Nutshell LarKC goes beyond the current limited storage, querying and inference technology Fuse reasoning with search aims for a paradigm shift Achieve “Web Scale Reasoning” Health Care and Life Sciences Interest Group #9

10. Our Objectives Integrate information using RDF data model – Integrated data sources to cover the path: gene – proteins – pathways – targets – disease – drugs – patient Reason over the integrated dataset – Remove redundancy / generate new links – Derive new implicit knowledge (e.g., “caspase activation via cytochrome c” is special form of “apoptosis regulation”) Apply information extraction algorithms and generate semantic annotations – Perform named-entity recognition and analyze some of the literals (e.g, the document texts) – Validate the new relations with the structured information Do it on a very large scale! Health Care and Life Sciences Interest Group #10

11. Semantic Annotation Generation Semantic annotations stands for: – Textual references to formally described entity in ontology (e.g., LLD) – The process of semantic annotations generation – Closes the gap between the structured and unstructured knowledge Common tasks related to semantic annotations – Automatic semantic annotations of text – Ontology population – Semantic indexing and retrieval of content – Query and navigation involving structured knowledge – Combination with classical information retrieval tasks Health Care and Life Sciences Interest Group #11