Finding the Best Molecule for the Job
Computer modeling may help narrow the molecular landscape to the best drug prospects
Every pharmaceutical company wants to find the next blockbuster drug. Yet finding molecules with a complete set of desired properties is tricky because of the astronomical number of medium-sized organic molecules. Now researchers at Duke University have developed a novel way to design virtual molecules from scratch. The work was published in the February 17, 2006, online issue of the Journal of the American Chemical Society.
“The biggest challenge in chemistry is being able to design molecules for particular purposes,” says Weitao Yang, PhD, a professor of chemistry at Duke University. “You can only do experiments on real molecules, but virtual techniques let you use non-real molecules to explore the molecular space.”
Yang along with colleague David Beratan, PhD, professor of chemistry, and post-doctoral fellows Mingling Wang, PhD, and Xiangqian Hu, PhD, developed an innovative approach. Rather than calculate properties of an enormous number of possible individual molecules, their framework approximates the properties over a continuous landscape in which the individual molecules lie. The model relies on knowledge of how atoms can be joined based on the energy relationships between nuclei and electrons in atoms. This narrows down the possible combinations and smoothes out discrete characteristics, such as atomic number, and thus provides a continuous surface for optimization.
For their proof of concept, the researchers focused on the properties that determine the ability of an atom’s electron cloud to be distorted by external electric fields. So, for example, if six groups of atoms could be located at each of two different sites, the model puts the different groups of atoms in the same spot simultaneously and then deter- mines how well the different combinations fit. This repeats at a predetermined number of sites. Joining the best molecular groups or combinations—like snapping together Legos—yields a complete molecule with the best properties.
This approach quickly yields the molecular potential, but it doesn’t necessarily map back to a molecule that can be made. For example, the best group at a particular site might be a combination of 13 percent of one molecule and 87 percent of another. This is impossible, of course, since only one molecule can occupy a single location, so the preferred molecule would be used.
“I think it’s very elegant how Beratan and Yang approached the problem,” says Ursula Rothlisberger, PhD, an associate professor of computer-aided inorganic chemistry at the Swiss Federal Institute of Technology in Lausanne, “But as a first step, it still has many limitations.” For example, it can only create simple molecules, as Yang would agree. He and his colleagues are now refining it to handle more complex systems such as designing optical materials for electronic devices. They plan to extend their work to drug design as well.
“We want to uncover many new materials that researchers didn’t know about before,” Yang says. “This method explores the design space much more efficiently.”