Using the Virtual Tumour to predict and optimize drug combination regimens
Eric Fernandez (1), David Orrell (1), Damien Cronier (2), Lawrence M. Gelbert (3), David A. Fell (1), Dinesh De Alwis (2) and Christophe Chassagnole (1)
(1) Physiomics plc, The Oxford Science Park, Oxford, OX4 4GA, United Kingdom ; (2) Lilly UK, Erl Wood Manor, Windlesham, GU20 6PH, United Kingdom ; (3) Lilly Research Laboratories, Lilly Corporate Center, Indianapolis, Indiana 46285, USA.
Objectives: Drug combination therapy is widely used to treat cancer because of the limited number of malignancies that can respond to single-agent chemotherapy. But when multiple drugs, combination schedules, sequences and doses are considered, the number of possibilities increases combinatorically, and can not be realistically tested either clinically or in animal models. We have designed a computational model, the Virtual Tumour, to help predict, optimize and allow prioritisation of the most effective drug combinations schedules.
Methods: The Virtual Tumour [1] is a computer model of a growing tumour, which integrates the cell division dynamic with the effect of antineoplastic agents. The simulation platform is composed of PK models of the drugs of interest, as well as a pharmacodynamic model of cell cycle progression. Drug effect can be calibrated by using various data sources: in vivo target inhibition (IVTI), xenograft growth timecourses, flow cytometry and public literature data. In order to validate the predictions made with the Virtual Tumour, we set up a validation study with our partner Lilly. The goal was to predict the xenograft course for two different combination regimens – one simultaneous and one sequential – using two drugs. The simulated timecourses were compared with experimental results in a single-blind test.
Results: We accurately predicted xenograft growth of two anti-cancer drug combinations using experimental data collected from single drug exposure uniquely. We show how a computational approach helps explain how multiple drug exposure and correct sequence leads to synergy, and how it can be used to subsequently design optimal schedule and combination treatments.
Conclusions: We have developed the Virtual Tumour model to aid with the design of optimal drug schedules. The model combines disparate data, at the cell and tumor level, into a consistent picture, and leverages them to make testable predictions about tumor response. Thousands of simulations can be performed if necessary to find the best treatment regime, reduce experimental costs and provide insights on optimizing use of a particular drug.
References:
[1] D. Orrell and E. Fernandez, Using Predictive Mathematical Models to Optimise the Scheduling of Anti-Cancer Drugs, Innovations in Pharmaceutical Technology, p59-62, June 2010.