INNOVATION | Fragment-based drug discovery is producing hits 

By Malorye A. Branca 
Senior Informatics Editor

April 16, 2004 | The lead cancer compound that Plexxikon delivered to Genentech last fall represents more than a nice milestone payment for a small startup not even three years old. Generated after less than six months of work, that new kinase inhibitor is a strong sign of maturity for a new approach to drug discovery.

"Fragment-based discovery is seeing traction," says Harren Jhoti, chief scientific officer of Astex Technology, another leader in this emerging field.

It's a big new idea that keys on very small molecules, usually referred to as fragments or scaffolds. The companies involved say they are turning around the risky business of lead identification, or finding starting points for drugs. They are not merely starting with more hits, says Michael Milburn, senior vice president of research at Plexxikon. "What we end up with, in terms of a lead compound, has a much higher probability of being successful than traditional hits." Jhoti concurs: "In a typical pharmaceutical company, 70 percent of the initial hits ultimately fail. [At Astex], 80 percent of the fragment hits are useful."

If it holds up, that will be welcome news for the entire industry. After a decade or more of watching high-throughput screening (HTS) systems churn out few good drug leads, most drug makers are desperate for a better approach.

"The hope was that HTS would generate hits with multiple chemotypes," Milburn says. Having more starting points to choose from should mitigate the risk of failure. However, there has been widespread dissatisfaction with the quality of the leads coming from HTS. Most still require much additional work — and often fail anyway.

"HTS has been disappointing," says Edward Roberts, chief scientific officer of Kemia and former head of discovery chemistry at F. Hoffmann-La Roche. "It yielded some good compounds for easy targets, but failed with most of the more difficult ones."

Difficult targets, of course, are the most valuable. Combinatorial chemistry, aimed at expanding compound diversity, has also fallen short. Researchers can make millions of compounds, but they still aren't producing the ones they need. "Some of these earlier technologies may not be as effective at generating hits as we had initially thought they would be," Jhoti says.

This problem has spurred heated debate about how best to mine chemical space. The number of biological targets in the body is limited, but a virtually unlimited number of potential chemical structures exist. So which kinds should drug developers pick? Most approved drugs have strikingly similar chemical make-ups. As a result, precious little is known about how to design different ones.

Many companies have simply stepped up the numbers game, throwing more and more molecules into their high-throughput screens. Others are trying to focus on selected parts of the druggable genome, such as the kinases, G-protein coupled receptors (GPCRs), or ion channel worlds. It should be easier to master them one group at a time, or so the theory goes.

Companies such as Astex Technology, Plexxikon, and Sunesis, meanwhile, are taking a novel tack. They start by scrutinizing the fit between the target and tiny "sub-small molecules." Because they are smaller, fragments don't stick as strongly as the larger molecules in traditional HTS libraries.

But the strength of that first connection isn't what matters most, say fragment-discovery proponents. The key is to get a snug fit, and then amass detailed information about how the fragment and target interact. From there, features can be added bit by bit, very selectively.

Starting small has several advantages, most notably that it reduces overall complexity. "As you go from hit through to preclinical development, a compound grows by 80 daltons on average," Jhoti says. By the time all the necessary chemical bells and whistles have been added to produce a functional and druglike molecule, many candidates have become dangerously large. Research shows that compounds over 500 daltons in weight have a much lower chance of reaching approval.

The fragments, Jhoti says, "start between about 100 to 250 daltons." Hence, researchers still have room to build them up without crossing the 500-dalton threshold. The goal is a molecule that still hooks onto the target at multiple points, but carries minimal excess chemical baggage.

According to Jhoti, the concept of keeping compounds small dates "back to the early '80s." So why is it only just starting to bear fruit? The answer, of course, is technology. For one thing, because the fragments bind weakly, they aren't revealed by traditional bioassay-based screens. Tools such as X-ray crystallography and nuclear magnetic resonance (NMR) are required.

The information is difficult to get, and researchers need a lot of it. "Until now, it's been much more laborious to get these data," Milburn says. But the necessary tools are finally becoming available. X-ray crystallography can be carried out at high throughput, and fragment-based NMR — focusing on only a portion of a molecule — is possible. The databases and modeling tools are also rising to meet the fragment-based designers' needs.

Novel biochemistry is also helpful for the subsequent tinkering. For example, Sunesis' new "tethering" technology temporarily tightens the binding between a well-fitted fragment and its target. That makes it easier to find the right fragments to start with and to add on additional chemical features.


Picking the Pathway 
A big question, Sunesis CEO Jim Young says, is: "What proteins and drug pathways do we work with?" Most pharmaceutical researchers are clustering around the same well-worn pathways. By bringing new chemical approaches into play, Sunesis can "expand the definition of what is druggable," he says. That's an important edge, which is winning the companies deals. One of Astex's major deals, with AstraZeneca, is for a "previously intractable target," Jhoti says. Astex received a milestone for that Alzheimer's disease-related project last fall.

All the compounds that have sprung from this new field are still moving along the pipelines, but the field is full of promise. Two-and-a-half-year-old Plexxikon is on track to file INDs for two internal projects by the end of 2005. Astex, which was founded in 1999, also expects to have two drugs in clinical trials by then. Things are going well enough that Astex has publicly discussed the possibility of an IPO. "We're watching the window," Jhoti admits.

Sunesis is getting similar rapid results. The company's kinase program delivered two "very interesting" inhibitors within only 14 months, Young says. "They exhibit the three most desirable hallmarks [of lead compounds]," he says: "Potency, selectivity, and druglike-ness."

Another encouraging sign is these companies' healthy deal portfolios. Astex recently signed up Boehringer Ingelheim, and already has agreements with other partners, including Schering AG, AstraZeneca, Aventis, Mitsubishi Pharma, and Fujisawa. Sunesis' partners are Merck, Johnson & Johnson, and Biogen Idec. And Plexxikon, the youngest of the three, already has that high-profile partnership with Genentech.

Fragment-based discovery "is unique and effective," Kemia's Roberts says. "But it's very important to choose the [initial binding] site very carefully," he cautions. Targets vary, and some pose unique challenges. Kemia, for example, is using a different novel chemical approach against only certain "hard-to-crack" GPCRs.

It's no panacea for productivity, but fragment-based discovery is looking increasingly appealing. "We can now get very detailed understanding of the forces that cause an inhibitor to bind with a particular region of a target," Young says. Now, if they can just put that information to optimal use, these rising companies might be able to grab a lead in the race to solve pharma's toughest targets.