QP-30

S. Ermakov,1 P. Forster,2 J. Pagidala,1 M. Miladinov,1 A. Wang,1 D. Bartlett,3 R. Baillie,3 M. Reed,3 T. Leil1; 1Bristol-Myers Squibb, Princeton, NJ, 2Forster Solutions, LLC, Wilmington, DE, 3Rosa & Co LLC, San Carlos, CA

BACKGROUND: Currently there is no single systems-level modeling software that ranks favorably against multiple criteria, e.g. model development capabilities, friendly user interface, export-import options, cost of ownership, etc. We developed Virtual Systems Pharmacology (ViSP) platform designed to easily set up and run multiple simulations in a flexible hardware/software environment.
METHODS: ViSP separates the instance of a simulation from the software that sets up the simulation. It relies on an executable file produced by compiling the model code into a binary file. The executable is initialized with a number of parameters, some are model specific, e.g. disease characteristics, properties of a particular patient or a drug, while others define simulation time, output frequency etc. Multiple executable files with different initial conditions can be initiated and run in parallel on separate processors, or in a cloud environment. This process was implemented in ViSP to handle executable files originating from different modeling software packages initialized with appropriate data. ViSP relies on web-based UI designed to be configurable by a user to accommodate the specifics of a particular model. It is capable of handling multiple models and large number of parameters presenting them in tree-like structure.
RESULTS: ViSP was used with a mechanistic metabolic diseases model to simulate the effects of metformin and a GPR40 agonist on glycemic biomarkers in type 2 diabetes patients. ViSP permitted easy set up of the clinical study design and simulations for a high dimensional model (> 100 ODEs, > 800 parameters) with multiple virtual patients.
CONCLUSION: Web-based user-friendly software was developed for running multiple simulations in a flexible hardware/software environment that is neither model nor software specific.