PharmML & SO - standards for encoding models and results in PMX and QSP
Maciej J Swat (1) and Nadia Terranova (2), on behalf of all DDMoRe contributors
(1) EMBL-EBI, Cambridge, UK (2) Merck Institute for Pharmacometrics, Merck Serono S.A., Switzerland.
Objectives: New formats enabling the efficient exchange and integration of pharmacometric (PMX) and quantitative system pharmacology (QSP) models across software tools have been defined and implemented as key elements of the DDMoRe interoperability platform [1]. Specifically, PharmML has been designed as the exchange medium for mathematical and statistical models [2, 3], and the Standard Output (SO) has been developed as a complementary component for storing typical output produced in a PMX workflow.
Methods: PharmML has been devised as a declarative language for mathematical and statistical models. Its development has been based on requirements provided by the DDMoRe community, popPK/PD and QSP partners, and on specific use cases from the main target tools (NONMEM, Monolix and BUGS). SO has been designed to be a tool-independent storage format providing a flexible structure for main results produced in PMX analyses. For this purpose, numerical results from target tools were collected, compared and classified in order to define and to implement a suitable structure able to account for typical output results. These two formats are subject to continued testing performed by a group of modellers and developers within DDMoRe, including academic and EFPIA partners. Standards extension to include additional features is ongoing.
Results: PharmML provides a structure for encoding continuous and discrete data models equipped with complex variability structure, covariate, structural and observation models. Definition of clinical trial design and modeling steps is possible as well. As a comprehensive self-contained format, it allows to encode models in tool agnostic manner [4]. SO is capable to capture any type of results from target tools including estimation, optimal design and clinical trial simulation tasks, thus, enabling effective data flow across tasks, and facilitating information retrieval for post-processing and reporting. These two formats facilitate (i) smooth and error-free transmission of models between tools, (ii) use of complex workflows via standardized model and output definitions, (iii) reproducibility of research, (iv) bug tracking and (v) development of new tools and methods.
Conclusions: PharmML and SO, as essential elements of the DDMoRe interoperability platform, proved to be capable to handle complex modeling scenarios, and to facilitate model exchange and results storage across various tools.
This work is on behalf of the DDMoRe project [1].
References:
[1] DDMoRe project, www.ddmore.eu
[2] Bonate, P. (2011). Pharmacokinetic-Pharmacodynamic Modeling and Simulation. 2nd ed.
[3] Lavielle, M. (2014). Mixed Effects Models for the Population Approach: Models, Tasks, Methods and Tools. Chapman & Hall/CRC Biostatistics Series.
[4] Swat et al. (2015). Pharmacometrics Markup Language (PharmML): Opening New Perspectives for Model Exchange in Drug Development. CPT Pharmacometrics Syst Pharmacol, 4(6):316-9.
