Measuring what makes a medicine

Researchers' method to rank molecules may aid in search for new drugs.

A new method for rating the attractiveness of a compound could help chemists discern potential new drugs from duds. Researchers have come up with a way to quantify a compound’s drug potential that moves beyond simply “hot or not,” instead providing a measure that allows compounds to be ranked as well.

COLOR ME DRUGLIKE In charting the “drug-likeness” of related compounds (represented as dots), a few tight clusters (left) indicate chemical similarity. Smaller, dispersed bunches and free-floating dots (right, for a different type of compound) indicate a lot of chemical variation. The redder the dots, the closer the compounds are to a QED value of 1, meaning they are the most druglike.  G.R. Bickerton et al/Nature Chemistry 2012

The approach “takes things a step further, looking at multiple factors instead of yes/no,” says chemical informaticist David Wild, of the Indiana University Bloomington,  who was not involved with the research.

The new technique uses eight molecular properties — such as the number of rotatable bonds a molecule has — that influence things like a compound’s toxic effects or its likelihood of being absorbed in the body. With some clever math, those probabilities are turned into a number between zero and one. When researchers tested their method against existing techniques for screening compounds, it outperformed the standard approaches at distinguishing known drugs from other molecules, the team reports in the February issue of Nature Chemistry.  

And because the new method, called QED, or quantitative estimate of drug-likeness, provides a numerical rating, it allows chemists to prioritize molecules for drug development, says study leader Andrew Hopkins, an expert in drug discovery and molecular design at the University of Dundee in Scotland.

Existing screening techniques are often used to make pass/fail judgments on compounds’ drug potential. Lipinski’s famous Rule of Five, for example, which uses measures such as a molecular mass not greater than 500 daltons to evaluate whether a compound might be absorbed and used by the body, has become a way to filter whole libraries of compounds even though it was just meant as a guideline, Hopkins says. This means potential drugs might be routinely screened out before they’re even given a chance.

Some chemists actually aim to break the rules, with the hope of finding a drug no one has bothered to look at, a sound approach given that 16 percent of today’s oral medicines —including some well-known drugs — violate at least one of Lipinski’s rules.

“Our metric suggests you can break some rules,” says Hopkins. “As with people, you can tolerate some bad behavior in someone’s personality if they are very good in their other qualities.”

In addition to assessing 771 oral drugs approved by the Food and Drug Administration, the researchers used QED to evaluate molecular properties of drug targets, the binding sites in the body that drugs latch onto. Since QED evaluates compounds on a continuum, it can reveal whether some targets’ chemical traits make them harder to get to than others, potentially highlighting the need for an innovative attack method.

QED also compared favorably with the gestalt assessment of chemists. Hopkins and his colleagues compared their technique’s evaluation of molecules with the opinions of 79 chemists who were asked whether they would pursue a potential compound. The QED values for both attractive and unattractive compounds were in line with the chemists’ ratings, suggesting the method nicely eyeballs a compound’s potential drug-worthiness.

“Chemists do have a concept of good, bad and ugly compounds,” says Hopkins.

The number of potential compounds and targets is far too large, however, for chemists to consider one-by-one. Perhaps QED can lend a hand, Wild says. “Chemists never like being told what to do by a computer, but at least the computer can help them test ideas.”

More Stories from Science News on Chemistry

From the Nature Index

Paid Content