KIWI: a collaborative platform for modeling and simulation
Sťbastien Bihorel, David Fox, Luann Phillips, Jill Fiedler-Kelly, Andrew Rokitka, Cindy Walawander, Ted Grasela
Cognigen Corporation, Buffalo NY
Objectives: Drug development programs rely increasingly on pharmacometric analysis to support decision-making and submissions to regulatory agencies[1,2]. To ensure high quality analysis, organizations must apply state-of-the-art science and implement a comprehensive infrastructure of procedures, workflows, and informatics capable of efficiently organizing, processing, maintaining, and communicating the volume of data and results generated by pharmacometric departments over the duration of a development program[3,4]. KIWI is a novel platform designed to meet these challenges.
Methods: A cloud-based application was developed as a model development platform for global teams to work and communicate results within a shared and consistently organized workspace. This validated platform allows 24/7 secure data access via web browser. It leverages the power of computational grid environments and facilitates the use of NONMEM, Perl-speaks-NONMEM, covariate search methods, and high-quality R-based diagnostic plotting.
Results: Users interact with KIWI through web modules. WORKBENCH organizes the workspace and implements a permission system defining user access to datasets and model results. It provides tools to manage datasets and control files and to submit jobs to a validated grid environment. MANAGE reports the list and history of completed runs and allows comparison of run results. SUMMARIZE automatically computes customizable summary statistics. VISUALIZE provides users with limited or no knowledge of R tools to create and view customizable diagnostic plots for estimation or visual predictive check runs. ANALYZE provides tools to facilitate covariate searches using stepwise forward/backward methods or generalized additive model analysis. All modules generate formatted outputs for export into technical documents.
KIWI accelerates modeling projects by decreasing the need for custom statistical or graphical code and eliminating quality checks of exported tables and plots. Finally, the validated workflow and tools promote traceability and reproducibility of results and reduce data manipulation errors.
Conclusions: As an integral part of an ongoing effort to enhance cross-department team collaboration and systematize the modeling and simulation workflow, KIWI was developed as an intuitive cloud-based platform for pharmacometrics designed to meet the demands of global teams.
 Grasela TH, et al. Pharmacometrics and the transition to model-based development. Clin Pharmacol Ther. 2007;82:137-142.
 Grasela TH and Slusser R. Improving productivity with model-based drug development: an enterprise perspective. Clin Pharmacol Ther. 2010;88:263-268.
 Grasela TH and Dement CW. Engineering a pharmacometrics enterprise. In: Ette EI, Williams PJ, eds. Pharmacometrics: The Science of Quantitative Pharmacology. Hoboken, NJ: John Wiley and Sons Inc., 2007:903-924.
 Grasela TH, et al. Informatics: the fuel for pharmacometric analysis. AAPS J. 2007: 9(5): Article 8. DOI: 10.1208/aapsj0901008.