Practical Implications of Ontology and Global Standards for Model-based Data Analysis
Chris Franklin (1), Bernard de Bono (2), Pierre Grenon (2), Sarala Wimalaratne (2), Oscar Della Pasqua (1)
(1) Clinical Pharmacology Modelling and Simulation, GlaxoSmithKline, Stockley Park West, Uxbridge, Middlesex UB11 1BT, (2) European Bioinformatics Institute, Wellcome Trust Genome Campus, Cambridge CB10 1SD, UK
Objectives: The development of a semantic metadata infrastructure is required to support the sharing and reuse of model libraries. The DDMore consortium is delivering an infrastructure based on the RICORDO approach to model annotation and data resources with shared ontologies. Ontologies consist in classes of entities found in a given domain as well as relationships between them. They can be used for the machine processable classification of information contained in model specifications, ensuring the shared clarity of these specifications. The approach enables the application of ontology-based knowledge management to model and data library. We illustrate the concepts with a real-life case study.
Methods: Models will be selected representing typical modelling approaches and languages drawing from examples of PK, PBPK and PKPD analyses. These will be inspected in relation to their parameters and variables (deterministic and stochastic components). Ontologies will be drawn from those in the public domain as well as the ongoing knowledge representation efforts from within the modelling community including the DDMoRe consortium. The results will allow comparable models to be used across different software applications and across different statistical methods.
Results: The case studies illustrate how the availability of standards for model specification permits improvement in model reusability on the basis of clear, explicit and shareable model descriptions in support of model libraries. The availability of specification standards for models coupled with standardised semantic annotation also implies that some models can be converted into a model definition language to provide interoperability. Furthermore these standards allow workflows for modelling and simulation within the DDMoRe framework.
Conclusions: Our work shows the need and relevance for community standards in model representation and model annotation for effective collaboration and workflow. The examples offer insights to the potential impact for both modellers and software development. This work shows that ontologies for model specification can support model reuse and establish crucial interoperability between models and experimental or clinical trial data (e.g. CDISC).
References:
[1] The RICORDO approach to semantic interoperability for biomedical data and models: strategy, standards and solutions, Bernard de Bono, Robert Hoehndorf , Sarala Wimalaratne, George Gkoutos and Pierre Grenon
[2] The DDMoRe Consortium http://www.ddmore.eu
