Standardized Output: flexible and tool-independent storage format of typical M&S results
Nadia Terranova (1), Marc Lavielle (2), Mike K. Smith (3), Emmanuelle Comets (4,5), Kajsa Harling (6), Rikard Nordgren (6), Duncan Edwards (7), Andrew C. Hooker (6), France Mentre (4,5), Maciej J. Swat (8)
(1) Merck Institute for Pharmacometrics, Merck Serono S.A., Lausanne, Switzerland, (2) Inria Saclay, France, (3) Global Clinical Pharmacology, Pfizer, Sandwich, UK, (4) INSERM, IAME, UMR 1137, Paris, France, (5) Univ Paris Diderot, IAME, UMR 1137, Sorbonne Paris Cité, Paris, France, (6) Department of Pharmaceutical Biosciences, Uppsala University, Uppsala, Sweden, (7) Simcyp (a Certara company), Sheffield, UK, (8) EMBL-European Bioinformatics Institute, Hinxton, Cambridgeshire, UK
Objectives:
The definition and implementation of formats enabling a reliable exchange of pharmacometric models across software tools is one of the key goals for efficiently promoting collaborative drug and disease modeling and simulation (M&S) research.
PharmML, one of the DDMoRe interoperability platform elements, has been designed to play the role of the exchange medium for models [1]. Similarly, the Standardized Output (SO) has been developed as a complementary element, tool-independent format, for storing typical output produced in a pharmacometric workflow.
Methods:
Based on the requirements provided by the DDMoRe community, SO represents a flexible storage format for typical results and information coming from a pharmacometric step, performed in any DDMoRe target tool (e.g., NONMEM, MONOLIX, winBUGS).
The standard is developed as an XML schema definition, which reuses some structure elements of the PharmML schema [2] (e.g., for the specification of matrices and datasets) as well as some other existing standards (e.g., UncertML [3] is used to encode uncertainty).
Results:
In the current version the SO structure consists of the following seven main sections:
- Tool settings: storing the reference to any file containing the tool settings of the performed task;
- Raw results: storing the reference to original output files produced by the target tool;
- Task Information: holding information about the modelling step execution (e.g., tool message, execution time);
- Estimation: storing estimation results produced by different estimation techniques (e.g., MLE, Bayesian);
- Model Diagnostic: storing information resulting from typical model diagnostic plots (e.g., individual fits, VPC);
- Simulation: storing simulation results (e.g., individual time course, population and individual parameters, random effects, covariates, dosing records) for each replicate;
- Optimal design: storing results (e.g., parameters values and precision, criteria, tests) obtained from an evaluation/optimization step.
A set of converters has been developed to allow for SO elements population directly from target tools.
Conclusions:
Beside capturing and storing various type of information, as a generic output model, SO aims at enabling effective data flow across tasks, thus, extending the workflow capabilities, and support the user in assessing, reviewing and reporting a modelling step.
This work is on behalf of the DDMoRe project (www.ddmore.eu).
References:
[1] The DDMoRe project, www.ddmore.eu
[2] Maciej J Swat, Sarala Wimalaratne, Niels Rode Kristensen, Florent Yvon, Stuart Moodie, Nicolas Le Novere, January 2015. Pharmacometrics Markup Language (PharmML) – Language Specification, Version 0.6. Available at PharmML website http://pharmml.org
[3] Uncertainty Markup Language: UncertML Version 3.0. Available at http://www.uncertml.org.
