Making drug development less of a gamble

Posted on 04. Oct, 2011 by in Academic Departments, Annual Report, Chemical and Biological Engineering, Issues, Research

The models Chemical and Biological Engineering Associate Professor Christos Maravelias develops are somewhat like a crystal ball that pharmaceutical companies can use to make research and development decisions about which drug formulations to develop.

Drug development is an expensive, highly risky, long-term endeavor. There might be hundreds of candidate compounds, for example, for just one “target” disease. Deciding which of those candidates to pursue involves extensive testing, time, and personnel and financial investments.

Maravelias develops methods that incorporate a company’s desired level of risk and account for all of this uncertainty. “From a portfolio of potential drugs, we try to develop methods that would allow us to select which ones to further develop, how to prioritize and also how to plan for our resources,” he says.

In drug development, the uncertainty is rooted in decisions people make. These decisions result in challenging optimization problems. “The underlying optimization problem is very interesting because of the uncertainty and the way it depends on what we are doing,” says Maravelias. “It’s a hard problem that hasn’t been studied in the literature.”

In a general optimization problem, there is a set of constraints that describes the problem and a set of optimization decisions. “In stochastic programming, it’s the same idea, but we generate multiple scenarios describing different uncertainty realizations,” says Maravelias. “We have to make the decisions that would be good in all these scenarios.”

This multiple-scenario approach leads to very large computational models, and another aspect of Maravelias’ work is developing the theory and solution methods—algorithms—that can efficiently solve the models for real-world applications.

Maravelias, who in 2006 earned a National Science Foundation CAREER Award for this research, says the models also could extend into many industries in which businesses are developing new products with limited resources.

Tags: , , ,

No comments.

Leave a Reply

E-mail address