Interactive Simulation and Visualization of Drug/Disease Models
Michael Heathman (1), Donald Jennings (1), and Bernett Lee (2)
(1) Eli Lilly and Company, Indianapolis, Indiana, USA; (2) Lilly Singapore Center for Drug Discovery, Republic of Singapore
Objectives: Drug/disease modeling enables quantitative decision making in the drug development process, through codification of scientific information about disease states, comparators, and new molecular entities. Model-based simulations are a powerful tool to leverage this information, facilitating open discussion with clinicians and other key members of the drug-development team. Unfortunately, the ability to answer clinically relevant questions is impeded by the complexity of the simulation process, and the long turnaround time required to conduct large-scale simulations.
To address this problem we have undertaken the development of an interactive system for simulation and visualization of drug/disease models.
Methods: The system was designed to include the following capabilities:
1. A standardized model-specification language, compatible with mixed effect models developed in common software packages.
2. An interface which allows user input of model parameters, drug properties, and patient characteristics.
3. A real-time simulation engine, designed to generate virtual patients and/or study data in a parallel fashion on a cluster of networked computers.
4. An interface for visualization of simulation results, calculation of summary statistics, and output of virtual patients for further analysis in other statistical software.
Results: An interactive environment for simulation and visualization from drug/disease models has been implemented at Lilly. Model specification is performed using a library of standardized R functions, while a graphical interface allows specification of simulation parameters and job submission to the simulation engine. A custom interface for visualization of these results has been developed using TIBCO Spotfire®, allowing quick and easy analysis of simulation output.
Conclusions: This system achieves two key objectives: (1) generation of simulation results on timescales that support real-time collaborative analysis, and (2) expansion of simulation capability to a broader non-technical audience for increased exploration of drug/disease models. The resulting knowledge supports decision-making related to compound selection, dose selection, and study design optimization.
