Computer cracks crystal-structure challenge
Program that predicts shapes of drugs may help companies save cash.
A man-versus-machine challenge has proven that a computer program can help crack the problems that pharmaceutical companies must tackle when marketing new drugs.
The computer program that met the challenge is that of theoretical chemist Sally Price of University College London and her team; it has already attracted the attention of drug companies keen to put the program to work.
The problem such companies face involves the fact that useful drugs often take several shapes or conformations, known as 'polymorphs'. And although chemically the same, a drug with a different crystal shape can have different properties, for example, it can take longer to dissolve in the bloodstream.
This means drug developers can only patent and market approved polymorphs, and they must ensure that their production process carefully produces only the correctly shaped drug.
They had trouble working out what had gone wrong and getting their original polymorph back. In the end, they had to gain approval for the new polymorph, which was less soluble in the bloodstream and, as a gel capsule, had to be refrigerated. This is estimated to have cost them hundreds of millions of dollars.
To anticipate such potential problems, researchers attempt to find out about the polymorphs of drugs using lab techniques such as crystallography. But this takes an enormous amount of time and leaves a considerable amount to chance. Price's software aims to cut short this process and predict the shapes a drug might take.
The test for Price's team came when Colin Pulham, a chemist at the University of Edinburgh, UK, challenged it to predict a new polymorph of piracetam, a drug used to treat Alzheimer's disease. Pulham had just discovered a new conformation of this drug in his lab, and wanted to see if Price's team could guess its shape.
Price's team, which was already looking at piracetam as a test case for their software, fed all the information about the bonds between atoms of piracetam into their program. Using the processing power of 30 computers, the program considered some 50,000 possible crystal structures of the compound, with each structure taking 1-2 hours to generate and analyse.
After a few months, Price handed Pulham a list of possible polymorphs, ranked in descending order of probability. The conformation at the top of the list exactly matched the shape that he and his colleagues had found.
The team members, who present their work this week at the e-Science All Hands Meeting in Nottingham, UK, says the test proves the efficacy of their computerized method. "Being able to meet this blind challenge adds a lot of credibility," says Price.
It will be a while yet before companies can put their faith into Price's program. "It's not going to be superseding the experimental approaches any time soon," says Adam Matzger, a chemist at the University of Michigan in Ann Arbor, Michigan. But after further testing, it could cut down the work needed to isolate the best conformation for market, or to anticipate potential problems in production.
"Pharmaceutical companies would like nothing more than to be able to predict all the obtainable polymorphs of a compound so they could patent them," says Matzger.
Price says that drug companies have expressed a great deal of interest in her lab's work but her team has yet to reap the rewards of meeting the initial challenge from Pulham. "Colin did say he'd buy us drinks but that hasn't happened yet," she says.