Send to printer »

Feature Articles : Apr 15, 2009 (Vol. 29, No. 8)

Capitalizing on Fragment-Based Discovery

Computational and Medicinal Chemists Are Joining Forces to Turn Hits into Leads
  • Kathy Liszewski

Fragment-based drug discovery is gaining momentum as an efficient strategy to find and optimize leads. The technology first identifies small chemical fragments that bind weakly, yet efficiently, to the target, and then grows or combines them to produce a lead with a higher affinity.

Fragment-based discovery differs from high-throughput screening. The latter assesses libraries with up to millions of compounds bearing molecular weights of around 500 daltons and nanomolar binding affinities. In contrast, the early phase fragment-based lead discovery screens libraries with a few thousand compounds with molecular weights of ~200. The hits that have comparably low binding affinities are then optimized.
At the recent CHI conference on fragment-based drug discovery, participants discussed challenges and new insights into this emerging technology.

“To understand the concept of fragment-based drug design, think of a drug molecule as composed of building blocks like Legos™,” recommended Eric Springman, Ph.D., head of discovery, Locus Pharmaceuticals. “Pieces are connected to build up the compound. This is in antithesis to high-throughput screening, where large libraries of elaborate compounds are scoured to find a starting point for drug discovery.

“There are important advantages to fragment-based design. First, because you start with something that is less complex, you can effectively screen a broader swath of chemical space and make the final product more efficient. Second, experience from combinatorial chemistry and high-throughput screening has taught us to emphasize certain drug characteristics early in the discovery process. In fragment based-design, you can choose these characteristics from the very beginning of the process.”

High-Performance Computing

Locus conducts contract drug discovery research and also has internal discovery programs targeting protein kinases in cancer and inflammation.

“We use our computational platform to model proteins and to simulate fragment binding,” added Dr. Springman. “This requires a solid core of high performance computing to handle and mine the vast data sets we generate. One cornerstone of our platform is based on patented methods for estimating the free energy of binding for virtual drug fragments that are selected from an in-house database of ~15,000.

“The power of virtual fragment simulations is in the combinatorial search of chemical space. For instance, if we are looking for a molecule composed of three fragments and we simulate 2,000–5,000 fragments, we’re searching chemical space equivalent to 8–125 billion possible combinations. That’s 1,000-fold larger than a typical chemical library."

Dr. Springman provided a perspective on this new and emerging field. “Fragment-based approaches have led to a paradigm shift in the field of drug discovery. The fragment-based approach has enabled a much more close and productive relationship between computational chemists and medicinal chemists.

“In the past, there has often been a large gulf between the two, with each having different perspectives on the process of drug design. Fragment-based methods click with the way medicinal chemists think and the process is necessarily prospective in nature. This brings the two sides together to decide what to focus on and how to construct it. Ultimately, it is leading to an improved way to produce drugs.”

Cutting-Edge Applications

Although fragment-based screening has recently emerged as a synergistic approach to conventional in vitro or cell-based screens, there are a number of hurdles that must be overcome for the successful use of fragment-based discovery, advised Alex Kiselyov, president of deCODE Chemistry. “First, you need to know how robust your assay is. Incorporating multiple diverse checkpoints would help in assessing how accurate it is and how much you can trust your data. Running alternative assays is always a good idea as these provide for the independent assessment of data quality.”

A well-defined flow for developing drug candidates already has been established within the industry, noted Dr. Kiselyov. “At the earlier stages, one needs to make sure that the identified actives are biochemically sound and specific. As a next step, functional activity of hits or their optimized analogues is assessed in cell-based systems followed by proof-of-concept in vivo efficacy studies.

“In addition, a multitude of parameters need to be balanced including pharmacokinetics, metabolic stability, toxicity, and others. Many groups in the industry, including ours, have consistently found that a smaller lead-like hit provides a better starting point for a medicinal chemistry effort.”

deCODE is focusing on fragment-based protein crystallography. “We find that this is a highly effective approach to rapidly identify both novel ligands and alternative binding modes,” Dr. Kiselyov continues. “This approach, for example, allows us to work on difficult targets, including allosteric modulators and protein-protein interactions.

“The key to finding promising actives is our in-house Fragments of Life™ (FOL) library. This selection is compact (~1,300 molecules) yet chemically diverse. It contains molecules found in the cellular environment, their metabolites, as well as compounds that mimic protein architecture. In a sense, these are ‘molecular rulers’ for assessing binding areas within a target.

As the field progresses, new applications for fragment-based drug discovery are emerging. Andreas Kuglstatter, Ph.D., head  of protein crystallography, Roche, reported, “Once you have found fragments that bind to your target protein, there are many things to do beyond growing or merging them to increase potency. Screening a large compound library for molecules similar to the identified fragments allows you to rapidly discover better molecules.

“For example, we were able to improve the activity of a kinase inhibitor by 100-fold in less than three weeks. You can’t do that if you have to synthesize new compounds. Another application is scaffold hopping, in which you replace the core of a known inhibitor with a fragment in order to develop a lead with different properties.”
Dr. Kuglstatter said that finding lead compounds with novel intellectual property (IP) can be challenging.

“High-throughput screening can find many potent hits, but the problem is IP space. Fragment screening is useful here, as well. We identified fragments with very distinct kinase binding motifs not described in the literature. We then made a custom library for each and tested their potency in a panel of 400 human kinases. So we now have highly attractive, off-the-shelf hits for many human kinases to jump start new projects.”

Fragments can also be used to explore protein flexibility and to enable rational selectivity design, noted Dr. Kuglstatter.

“One of the biggest issues with kinase inhibitors is selectivity. By cocrystallizing fragments with their targets, we identified unique protein conformations. This opened the door for the exciting application of selectivity design. Fragments can be optimized using structure-based approaches to bind stronger to your drug target and weaker to all other kinases. This translates into better in vivo safety. We have used these new applications for all small molecule programs where the target is a soluble protein.”

Protein Kinases

Protein kinases play key roles in intracellular signal transduction and, as such, are involved in a variety of disease states including oncology, metabolic, and immunological disorders. Because these targets are druggable, kinases have moved into focus as promising therapeutic drug targets. Proteros Fragments is targeting kinases among other relevant proteins with their fragment-based lead generation technology that employs cocrystallization of target and leads followed by structure-based optimization.

According to Gerhard Mueller, Ph.D., CSO and managing director, “In the area of kinases, we apply a conceptually novel retro-design strategy for which we have synthesized tailor-made fragments and specifically developed an assay technology for efficient fragment screening. The ability to engineer specific binding kinetic characteristics is the main advantage of this kinase inhibitor design approach. In addition, we intentionally avoid at the outset of a lead-finding program placing these seed fragments in the adenine-binding region, which usually translates into an improved IP position for the subsequent optimization campaign.”

Dr. Mueller says that most protein kinases in solution adopt a wide variety of different conformations, so inhibitors can be designed that show longer residence times on the target. “Different target conformations offer novel binding sites with unique topologies. Our strategy and fragment design concept that we pursue in-house is based heavily upon protein crystallography-derived structure information of the highest possible resolution. This provides a clear-cut assessment if the fragment of interest is indeed targeting the desired conformational state.

“The key element of the retro-design strategy is that distinct binding attributes can be constructed on the fragment state to create kinase inhibitors with slow dissociation rates. This translates into candidates with increased cellular and in vivo efficacy.”

Structurally Related Libraries

The fragment-based drug discovery approach of Polyphor combines screening of a general compound library and a structurally related fragment library. Daniel Obrecht, Ph.D., CSO, said, “Traditionally, one starts from fragment hits that are then optimized. Our approach takes advantage of the possibility to combine structural information from fragments and structurally-related final compounds.”

The company’s general library consists of 140 core structures from which 600 scaffolds are derived to populate a screening library of 35,000 compounds. “An important feature is that both low and high variation substituents are kept constant throughout the scaffolds. This allows one to establish an SAR already after the primary screening.”

Polyphor’s fragment library consists of more than 1,000 fragments with molecular weights ranging from 150–350 daltons that are derived both from the scaffolds and the substituents of the Polyphor screening library. Dr. Obrecht explained, “By screening both libraries, a more comprehensive SAR can be obtained, one which allows a more directed entry into subsequent optimization and ultimately shortens the early drug discovery process.”

Natural Scaffolds

“The drug discovery world has always been looking at natural products as a valuable source of bioactive molecules,” said Dmitry Genis, Ph.D., CEO, Asinex.

“It is remarkable to note that 33 percent of new chemical entities discovered between 1981 and 2002 are natural products (or related to them). However, natural product chemistry is challenging and biologists have to deal with complicated screening procedures.

“Since the 1980s most pharmaceutical companies have reduced their efforts in this area, and the advent of synthetic combinatorial chemistry has further contributed to this decline.”

Dr. Genis reported that times are changing. “The past two decades of high-throughput screening have shown that the random generation of compound libraries by a pure combinatorial chemistry approach has created millions of molecules that did not match the requirements of the drug discovery industry. Therefore, it makes sense to use natural product structures as starting points for new drug-like libraries.

“Natural products represent an additional source of diversity that cannot be found in synthetic combinatorial libraries. Our BioCore approach is a synergic compromise that solves the problematic issues of natural products and combinatorial libraries and, at the same time, allows the unification of the positive features of these two areas.

“BioCore is the key component of our BioDesign approach. The idea is to create compounds that contain highly privileged elements from natural products and known drugs. A BioCore is the combination of two heterocycles—one is aromatic and one is saturated, linked by a carbon-carbon spacer fragment. We strongly believe that following the BioCore framework allows chemists to create nature/lead-like scaffolds,” explained Dr. Genis.

Although the idea of incorporating natural product elements is not novel, the company has identified what it feels are under-represented areas of chemical space and implemented those into real libraries. Dr. Genis reports that, “Based on the BioDesign approach, we offer research tools based on bio-building blocks, bio-fragments, and lead-like sets, enabling easy hit-to-lead follow-up.”

As more medicinal and computational chemists incorporate fragment-based approaches into their discovery toolbox, the field is expected to make further progress in accelerating the pace of drug discovery.