The one-compound, one-target model of drug discovery has long since been debunked as not being an efficient methodology for drug discovery. As budgets tighten across the industry, pharma of every size is seeking to get the most bang for its buck. At the Cambridge Healthtech Institute’s “Structure Based Drug Design” conference to be held later this month, a number of presenters are going to discuss how their respective companies are approaching these current challenges from different perspectives.
Fragment-based Drug Discovery
Evotec (www.evotec.com) uses the combination of a high-quality fragment library with sensitive biochemical screening methods for the identification of weakly active fragment molecules as novel starting points for medicinal chemistry optimization. “The application of this technology to enzyme targets such as renin and various kinases is extremely powerful and can dramatically reduce the times traditionally seen for moving from target to preclinical development as well as generating novel chemical start points,” Mark Ashton, evp of business development, says.
The traditional approach for screening for candidate compounds is high-throughput screening. “The hit compound identified via this approach is of a similar molecular weight and size to that expected of the final drug, so the optimization of the hit compound entails the sequential removal and addition of appropriate functional groups aimed at increasing the potency and improving the pharmacokinetics of the compounds.”
This can be a lengthy process, reports Ashton. “The advantages of fragments are that they are about one-half the molecular weight and so additional functionality can be quickly added to rapidly improve the potency and PK properties. The fragment-based approach ensures early access to crystal structures of the fragments interacting with the biological target. Having the crystal structures provides valuable information on how the ligands interact with the target, the interactions of the molecular bonds, and what functionalities are most important. This allows us to quickly build improved molecules and bypass the usual lengthy process of trial and error in molecule design.”
Ashton explains that Evotec has successfully demonstrated the use of high-concentration biochemical screening for fragments using its single-molecule fluorescence correlation spectroscopy detection technologies to ensure high reproducibility and sensitivity. “We also use x-ray crystallography to determine the binding mode of active fragments,” says Ashton.
“Once the active fragment has been identified, Evotec uses fragment optimization together with unique capabilities to advance the fragments to preclinical development candidates with improved speed and attrition.”
At the conference, Evotec representatives will discuss how the company has improved on its approach for identifying novel compounds for interesting biological targets. “Big pharma is looking for earlier stage discovery and innovation—one of the big challenges facing pharma companies today is that they need to focus on further strengthening their pipelines,” Ashton notes. “Innovative approaches to achieving this need to be identified to complement the more traditional methods of drug discovery.”
Multiple Kinase Inhibitors
Prior to the completion of the Human Kinome, the mapping of all the kinases in the human genome, drug discovery around kinases was problematic, notes Sucha Sudarsanam, CSO of Emiliem (www.emiliem.com). “I can tell you from the standpoint of someone who has worked with kinases for a long time that, until about 2001, kinases were viewed with suspicion as possible drug targets because there were too many of them, and there were also issues of selectivity.”
What Sudarsanam, and many others, have since discovered is that, when it comes to looking for efficacious compounds to take to the clinic for treating cancer, hitting multiple targets is actually a good thing. “Cancer is heterogeneous,” Sudarsanam says. “Each kind is involved with multiple pathways, angiogenesis, tumor growth, and metastasis, so hitting one pathway isn’t going to be enough.”
Computational techniques have provided a start for designing inhibitors against individual targets; however, designing compounds that are effective against multiple kinases has remained a challenge.
“Since we know so much more about the chemical space of kinases from published scientific and patent literature, it doesn’t make sense to start with high-throughput screening,” Sudarsanam adds. “What we use is a process called K-STAR, which is Kinase Structure-Target Activity Relationship. The basic idea is that we target the compound against multiple kinases.”
By starting with validated targets and known kinase inhibitors, Sudarsanam notes, you eliminate the hit-to-lead stage of drug development up to two to three years. “You save both in time and infrastructure cost up front,” Sudarsanam says. “This is truly translational medicine in that we look at these pathways closely, and by doing so, we can run focused clinical trials and can be specific about the cancer and patients we are looking to treat. It has been exciting.”
Protein and Ligand Structure-based Design for GPCRs
Protocols for structure-based drug design with soluble proteins are now fairly routine. For GCPR targets, however, structure-based approaches are likely to be limited to homology models for a few more years. “But we are on the verge of conquering this frontier,” says Sid Topiol, associate director for computational chemistry at Lundbeck Research USA (www.lundbeckusa.com), which is focused on central nervous system indications, primarily CNS activity surrounding anxiety and depression.
Topiol will present on three areas in his talk: examining traditional ligand-based-drug design; target design using structural and computational analysis—or taking advantage of what is already known about the system; and ligand redesign using structural information about proteins to tune out unwanted activity.
“We’ve used an arsenal of approaches over the last few years,” says Topiol. “The key is to take advantage of what we already know, what information about the ligands has already been published, and available information about crystal structures. In the area of GPCRs, we generally have sequence information but no detailed protein structural information, which limits most efforts to use rodopsin as a template for homology models.”
Topiol explains that on the drug side, depending on the target, there is often valuable information already in the public domain. “In terms of putting all this together, we try to capitalize on the hundreds of thousands of small molecule structures, to extract the information that might help us build in the properties in which we’re interested,” says Topiol.
The methods they use include studying structure-activity relationships and high-throughput docking, “which is the computational counterpart to high-throughput screening,” adds Topiol. “We can do this with large numbers of real or virtual compounds to sift out what interacts most favorably with a target of interest.”
Topiol reports that Lundbeck Research is eager to usher in structure-based drug design for GPCRs. “The design of the computational protocol is important in trying to answer the questions you need to address,” he adds. “Since we have limited in-house software development, we establish partnerships with academia and software vendors.”
Advances in Lead Optimization
Lead optimization has made huge advances over the last few years. “Of course, lead optimization is critical in drug design,” says Woody Sherman, director of applications science at Schrodinger (www.schrodinger.com). “Computational structure-based drug design is an area that has been making significant strides in the last few years, and our most recent developments in structure-based lead optimization put us in a place where we can start to accurately predict the ranking of compounds in a lead optimization series.”
Sherman will talk about the recent advances in lead optimization and some of the ways in which lead optimization by way of computational approaches can be more feasible and acceptable. “One of the chief obstacles facing computational approaches is accounting for receptor flexibility as you make changes to a lead series,” says Sherman. “When you approach the problem from a physics-based perspective, it is important that you treat the entire system accurately. We are firm believers in getting the science right.”
While Schrodinger was founded on quantum mechanics, it has been using varying flavors of that technology, along with other algorithms, to advance lead optimization. The most recent advances in the company’s Glide XP scoring function have allowed for the accurate prediction of binding energies of diverse compounds, which has not been possible with docking programs to date.
Sherman will also discuss rank ordering of congeneric molecules with an emphasis on using accurate charges derived from quantum mechanics. “Our goal is to make drug discovery development more efficient,” says Sherman. “However, one of the difficulties for people developing computational methods is that they don’t have the facilities to actually do the experiments needed to test many of their predictions, and most of the valuable data from companies is proprietary.”
There have been a fair amount of retrospective studies done, Sherman notes, but the results can be hard to assess because it is not clear how much tweaking of the parameters is needed to get the results presented. “Our field tends to spend a lot of time trying to predict existing experimental data, and in the end, the results for almost any method come out looking pretty good on paper,” says Sherman. “But our goal is to get the computational methods accurate enough to really add value to the discovery process. The ultimate proof of the methodology is when we hear from customers in biotech and pharma that their computational predictions were experimentally validated, and that has been happening much more frequently in recent years.
“The process of structure-based drug design has, at its core, the accurate prediction of structural complexes and accurate scoring,” Sherman adds. “If we can deliver the tools to make accurate predictions of binding energies in a reasonable amount of time given the computational resources available, we’re doing what we’ve set out to do.”
Novartis’(www.novartis.com) Aliskiren is the first new type of medicine in the management of hypertension, notes N. Claude Cohen, president and CEO of Synergix (www.synergix.com). “Pharmaceutical companies have been working for more than 20 years on this new therapy; it is an area of extreme importance because 50 percent of people affected by hypertension are not adequately controlled with current medication,” says Cohen. “Currently available drugs can help, but there are side effects and complications.”
Cohen has researched this area since the late 1980s. “Back in the early days, peptide molecules were thought to be the mode of delivery,” notes Cohen. “And that was a mistake, since peptides are rapidly metabolized. However, peptides do bind to proteins, so we had to develop a nonpeptide that mimics all the properties of the peptide.”
For this purpose, a homology model of the enzyme was used to characterize the binding mode of a peptide compound, CGP38560, in complex with a model of renin. “We didn’t have an experimental complex between the lead compound and the target, so we developed a model using computer simulation,” says Cohen.
The medical model they used in development of the compound, explains Cohen, is the renin-angiotensin cascade leading to hypertension, which occurs in three levels. “Most drugs in this class can block the cascade at the second and third levels, but the goal was to successfully block the cascade at the first level,” he adds.
Cohen confirms that structure-based drug design played a key role in the process of developing the compound. “It exploits the recognition and discrimination capabilities of the target protein to create favorable interactions in three dimensions with the molecule you conceive.
“To do so, it is necessary to have good knowledge of the 3-D structure of a complex with the protein, which we didn’t. That was where the computer simulations came in,” Cohen says. “And management was surprised when we managed to do it successfully.”
Consequently, four chemically unrelated nonpeptide series were discovered acting as renin inhibitors at the one through three nonmolar level. “We selected one of these leads for further development, and it led to Aliskiren, which was just approved by the FDA in March,” reports Cohen. “The hope is that it will be safe, useful, and efficacious, and that it will work really well with people who don’t currently respond to medication such as the elderly and diabetics.”
“The work of structure-based drug design is to discover the key to the lock,” Cohen concludes. “Disease targets are three dimensional locks, and structure-based drug design is now a mature discipline.”