Development and integration of the WinBUGS connector in the DDMoRe Interoperability Framework
Cristiana Larizza (1), Elisa Borella (1), Lorenzo Pasotti (1), Gareth Smith (2), Richard Kaye (3), Paolo Magni (1)
(1) Department of Electrical, Computer and Biomedical Engineering, University of Pavia, via Ferrata 5, Pavia, I-27100, Italy; (3) Cyprotex Discovery Ltd; (4) Mango Solutions.
Objectives: The objective of our work is to develop and test a connector integrating WinBUGS [1] in the DDMoRe Interoperability Framework platform. The connector allows the user to perform all the steps of a WinBUGS model execution (PharmML into WinBUGS model translation, WinBUGS run, generation and retrieval of the desired output) within the DDMoRe Interoperability Framework platform.
Methods: Connector development includes five steps: 1) Upgrade of our previously developed PharmML-to-WinBUGS model and NMTRAN-to-BUGS data file translation tools [2] to include new features (i.e., Bayesian priors) available in PharmML 0.7 and successive versions [3]. 2) Support to categorical covariates, piecewise function definition and correlation of random effects, which were not included in the previous version of the translator. 3) Implementation of a Java-based tool for standardized output (SO) file creation, summarizing the outputs of a WinBUGS run (CODA files), using lib-PharmML-SO [4]. 4) Development of a connector, via different shell scripts, that is responsible for: preparing inputs, invoking the tool, monitoring the progress of the processing and retrieving results from execution. 5) Definition of interoperability commands in the form of R functions to define the number of Markov chains, the number of total iterations per chain, the length of burn-in and other common options in Bayesian analysis. PharmML 0.8, WinBugs 1.4.3, BlackBox 1.5, PKPD Model Library 1.2, the WBDiff and WBDev interfaces and Java Libraries available in [5] were used.
Results: The connector was successfully tested on a variety of algebraic/ODEs single-subject/population models publicly available on the DDMoRe Model Repository [6]. For testing purposes, priors were properly added when not present to enable Bayesian estimation.
Conclusions: The connector allows the user to specify the quantities to be monitored, customize the MCMC sampling algorithm, run the execution, retrieve the standardized output, and perform graphical convergence diagnostics and posterior inference for a large number of modelling situations within the DDMoRe Interoperability Framework. Together with the Model Repository, it promotes the exchange and reusability of models in Bayesian framework, which could greatly improve drug development.
Acknowledgements: This work was supported by the DDMoRe project (www.ddmore.eu).
References:
[1] www.mrc-bsu.cam.ac.uk/software/bugs/the-bugs-project-winbugs/
[2] PAGE 24 (2015) Abstr 3565 [www.page-meeting.org/?abstract=3565]
[3] Swat MJ et al. Pharmacometrics Markup Language (PharmML): Opening New Perspectives for Model Exchange in Drug Development. CPT Pharmacometrics Syst. Pharmacol. (2015) 4, 316-319.
[4] PAGE 24 (2015) Abstr 3599 [www.page-meeting.org/?abstract=3599]
[5] sourceforge.net/p/ccopi/code/ci/master/tree/
[6] repository.ddmore.eu
