By Matt Luchette
August 21, 2012 | Beauty may be in the eye of the beholder, but Andrew Hopkins, professor of medicinal informatics and translational biology at the University of Dundee, Scotland, thinks chemical beauty is more objective than that.
In a landmark Nature Chemistry article published last January, Hopkins described his “Quantitative Estimate of Drug-Likeness” (QED), an algorithm to rank chemical compounds based on their oral bioavailability, helping pharmaceutical companies decrease the risk of developing compounds that will fail during FDA testing.
“Pattern recognition is the forte of the chemist,” Hopkins wrote. But he hopes the QED will be able to quantify that ability, adding that the measure “captures the abstract notion of aesthetics in medicinal chemistry.”
Similar metrics, such as Christopher Lipinski’s famous Rule of Five, have already been developed for the purpose of predicting a compound’s oral bioavailability and help guide drug companies in screening small molecule libraries for drug development. But the limitation of current methods, Hopkins argues, is that all compounds are seen as equally favorable or unfavorable. There is no way to compare different compounds to decide which may be “best” for certain applications. Rather, researchers should “think about druggability as a probabilistic matter,” Hopkins said.
The QED, on the other hand, is based on assigning a “desirability” ranking for each compound based on eight criteria, such as molecular weight and number of hydrogen bond donors, that have been experimentally determined to be important for a drug’s oral bioavailability. The rankings are based on a library of 771 approved orally dosed drugs. The more common a certain trait is in the library, the higher the desirability ranking. This ranking could help drug developers reduce the hundreds of millions of dollars lost from compounds that fail during human testing.
But Hopkins does not see the QED as being limited to just the pharmaceutical industry. The principles the algorithm uses to rank compounds based on a set of quantifiable criteria could be applied to any industry that needs to optimize chemical compounds. In this way, Hopkins argues that the QED is able to quantify a chemist’s intuition about the relative “beauty” of certain compounds, or how desirable certain compounds are for various chemical applications.
Other scientists have taken note of this quantitative power in the QED. Matt Segall, CEO of Optibrium, a British biotechnology company, is trying to modify the QED to make it even more useful for pharmaceutical companies.
“The QED is great because it gets rid of the hard cut-offs that Lipinski uses,” Segall told Bio-IT Worlds, but he feels Hopkins’s metric falls short by assigning high desirability only to traits that are more common in approved drugs.
Segall argues that certain traits may be more common simply because more research has been done on those types of compounds. Instead, he suggests the rankings should be assigned using Bayesian probability. “Bayesian probability theory allows us to quantitatively estimate… the probability of identifying a drug, not simply a value that is similar to known drugs,” he said.
Yet with so few data points in these less-researched areas of chemistry, one could argue that drug developers may end up making unfavorable design decisions in choosing these less popular traits. But Segall has confidence in the data. “We can’t make perfect predictions. We need to think about what decisions we can make with confidence without throwing away potentially good compounds,” he said.
Segall, like Hopkins, thinks the quantitative power in these models can ultimately help chemical and biological engineers make better design decisions. “What we do in biology is not design,” Segall said. “When Boeing designed their 777 aircraft, it flew the first time.”
But in the pharmaceutical industry, where it costs hundreds of millions of dollars and a decade or more to bring a drug to market with marginal rates of success, biological engineers are nowhere near that type of precision. By continuing to develop better algorithms for screening drug libraries, researchers like Hopkins and Segall stand to transform the biotech and pharmaceutical industry by using data to motivate design.
Matt Luchette is a Bio-IT World Contributor.