March 15, 2005 (Vol. 25, No. 6)

Definition Is Broadening: Not Just Structure-Based Design Anymore

Structure-based approaches for drug design and virtual screening give new meaning to Louis Pasteur’s saying “Chance favors the prepared mind.” In silico methods are becoming more efficient in allowing scientists to hone in on and manipulate specific molecular structures of interest. Why might this be preferable to the standard drug discovery process?

“What is the cost of setting up an experimental high throughput screening (HTS) laboratory and developing assays to discover a compound, compared to doing the same discovery with software?” Osman F. Gner, Ph.D., executive director, cheminformatics and rational drug design at Accelrys (San Diego), asks.

“Some estimate HTS starts at approximately $1 million per project, with no guarantee you’ll find anything useful. Comparatively, in silico methods can focus efforts, save time, and provide substantial savings.”

According to Dr. Gner, the advantage of using rational design is illustrated by the recent case of two companies that discovered the same small molecule by different means.

“A small molecule that blocks type I TGF receptor kinase (which is implicated in diseases such as fibrosis and cancer) was discovered independently by two companies, Biogen Idec (Cambridge, MA) and Eli Lilly (Indianapolis).

“While Eli Lilly ran a lengthy and expensive experimental HTS technology, Biogen performed a relatively quicker, less costly virtual screening. So, structure-guided design can be a powerful shortcut to years of painstaking benchwork.”

Dr. Gner also points to Crixivan, the top-selling AIDS drug of Merck & Co. (Whitehouse Station, NJ), which was generated by virtual methods. “Today, any company pursuing drug discovery and drug design must have a computer-aided design lab.”

Accelrys provides many tools for drug discovery such as its Catalyst software environment for high throughput screening of potential leads using 3-D searching tools, or docking and scoring tools such as LigandFit and Affinity.

Shape and Electrostatics

“Our mantra is that the chemistry of molecular interactions boils down to shape and electrostatics,” says Matthew Stahl, Ph.D., senior vp at OpenEye Scientific Software (Santa Fe, NM).

“Molecular shape is a combinatorially nasty problem. It’s difficult to know what conformation a molecule will adopt. So, the quality of answers obtained by molecular modeling depends upon the accuracy of your models,” asserts Dr. Stahl.

“Small molecule conformer generation is a key first step in structure- and ligand-based drug design. That’s why we have taken a meticulous approach by using state-of-the-art technology that generates coordinates from scratch for structures obtained in public databases. We then superimpose these structures and perform a variety of processes to see which conformers are reasonable. We can get 9599% of structures to match a minimal level of good fit. This is a significant number.”

According to Dr. Stahl, scanning for second-generation compounds could position companies to identify other novel bioactive structures. “For example, our software searches 3-D databases to look for different molecules with similar shapes to proprietary compounds.

“Even if the structures differ, as long as there are similar electrostatic properties, one can literally find hundreds of new bioactive structures. This type of virtual screening method is quick and allows discovery of a billion conformers in 24 hours.”

Open Eye offers several software programs for rational drug design. The newly released version (1.8) of its Omega software rapidly generates multiconformer databases suitable for large libraries required for computer-aided drug design. Also, the company’s ROCS (for Rapid Overlay of Chemical Structures) software performs large-scale 3-D database searches using a shape-based superposition method.

The spectrum of protein kinases, i.e., the kinome, consists of more than 500 proteins (mostly tyrosine and serine/threonine kinases). These play a leading role in signal transduction pathways and thus affect most fundamental cellular processes. Aberrant activities are associated with a number of diseases. Thus, kinase inhibition represents a way to dampen unwanted processes such as those involved in cancer.

Kinase Co-Inhibition

“The phosphorylation of proteins is a critical mechanism for cellular responses involving metabolism, energy, and signaling,” says Timothy P. Clackson, Ph.D., senior vp and CSO, Ariad Pharmaceuticals (Cambridge, MA).

“From a chemical point of view, most kinases share a similar architecture in their binding sites. Structure-guided design helps to localize subtle features and different conformations in the binding pocket. We can use this information to generate selective kinase inhibitors, but we can also choose to target more than one kinase at a time, i.e., co-inhibition.”

Dr. Clackson notes that the company has focused in the last several years on the need to inhibit multiple kinases simultaneously. “Co-inhibition is an attractive means to target cancers.

“For example, inhibiting Src and Abl kinases together has a greater effect against certain forms of cancer such as chronic myelogenous leukemia and Abl-dependent diseases. Structural biology modeling allows us to look for inhibitors by computational methods to enhance potency, solubility, molecular weight, and bioavailability.”

Inhibitory Profiling

Cyclin-dependent kinases play critical roles in regulating the cell cycle. Thus, inhibition of CDKs should arrest cell proliferation and trigger cell death in tumor cells.

AstraZeneca (Wilmington, DE) is utilizing structure-guided methods to design CDK4 inhibitors. According to David Buttar, Ph.D., scientist, some tumors have lost natural CDK4 control mechanisms.

Since CDK4 is unsolved at present, AstraZeneca’s approach is to exploit CDK2 crystallographic information as a surrogate for CDK4 design. Additionally, Dr. Buttar indicates that one way to look for structural surrogates is to classify kinases by their inhibitory profile.

“The past decade has witnessed an explosion in the number of three-dimensional protein small molecule structures from experimental as well as in silico approaches,” Juswinder Singh, associate director, computational drug design group at Biogen Idec, explains.

“The use of profiles enables a more empirical approach to leveraging the experimental data to more effectively filter these massive in silico datasets. The profile approach is broadly applicable to either specific drug targets or other gene families.”

According to Singh, “The gene family concept is emerging as a powerful paradigm to accelerate drug discovery through leveraging structural insights into druggability and transferability of chemotypes amongst related drug targets.

“The protein kinases exemplify a family where the potential exists to accelerate lead discovery and optimization by inferring between the massive amount of structural and chemical data from gene family members.

“We have developed novel tools to analyze protein kinase inhibitors and identified key interaction signatures that seem to be conserved across most of the kinase inhibitors that have been solved using x-ray crystallography. These signatures have been used by us to virtually screen for novel inhibitors.”

The concept of a kinase profile rather than specific kinase inhibition may provide a better method of obtaining a pharmacological response. Singh and colleagues have developed a profile-based structural interaction fingerprint technology.

The profile is a sequence position-specific scoring matrix encoding the probability of finding any of the 20 amino acids at a specific position in the target, notes Singh. The analysis derives a position-dependent profile of the probability that a given interaction is present at a specific site.

Singh believes that profile-based methods provide a superior enrichment of kinase inhibitors relative to traditional scoring approaches.

“At present, there are no effective tools to leverage the information gained from protein inhibitor complexes. This p-SIFt profile approach fills this capability gap, and provides a powerful mechanism to screen for novel inhibitors using virtual screening and also identify selective inhibitors.”

Custom Design Tools

Robert Jackson, CSO, Cyclacel (Dundee, U.K.), says that customizable software often provides the best approach to rational drug design. “Many generic software approaches are inefficient, therefore we decided to develop and utilize proprietary software to improve drug discovery and optimization. The beauty of having our own applications is that we can increase the efficiency of design by focusing in on specific interactions and areas important to binding partners revealed in x-ray, crystal, or NMR-derived structures.”

The company uses a proprietary docking program called Lidaeus, developed in collaboration with the University of Edinburgh. Campbell McInnes, Ph.D., head of the structure-based design group, says that the company has applied this docking program to identify several small molecule inhibitors of the human MDM2 (HDM2)/p53 interaction.

Normally, p53 remains in an inactive form by binding to HDM2 at its amino terminus. The company is investigating ways of disrupting that interaction in order to activate p53.

“Using Lidaeus to identify small molecule p53-HDM2 inhibitors and macromolecular NMR spectroscopy to study their binding, we observed significant changes and rearrangements in the HDM2 amino terminus that were not all localized to the inhibitor binding cleft.

“In order to characterize what this means for small molecule design, Cyclacel and our collaborators at the University of Edinburgh Biomolecular NMR group employed NMR to obtain the first atomic resolution solution structure of the HDM2 amino terminal regions in the absence of the p53 ligand or small molecules.

“Coupling that information with the known crystal structure of HDM2 complexed with a p53 peptide, we compared non-liganded and liganded structures. Using structure-based design, we are able to use this data to design inhibitors that more effectively targeted the p53 binding pocket on HDM2.”

New Trends

What’s ahead in the area of rational drug design? Bryan Koontz, vp of corporate development at Tripos (St. Louis), suggests that discovery informatics software for virtual high throughput screening (vHTS) will continue to play a vital role in rational drug design.

“Docking and scoring software are still important rational drug design tools for discovery scientists who are attempting to simulate what may happen in the lab.

“Other software tools, including programs capable of leveraging various scoring functions to invent new compounds with specific structural or physical properties, will also play an increasingly important role in discovery.”

Tripos recently acquired Optive Research (Austin, TX) and has integrated its EA-Inventor program in this area.

Other improvements will be ligand-preparation algorithms with improved structure representation capabilities. “Docking programs are often blamed for producing poor virtual HTS results. This may be due to the docking algorithm itself, but scientists should determine if they are docking the right structures to begin with,” Koontz says.

Often, this is not the case. Chemical compounds in nature may take various forms, depending on their natural environment. That is, there may be hundreds of different ways to represent that compound in a computer program. If a scientist only considers one formone structureof a compound that produces poor docking results, they may be inclined not to pursue that compound in the lab. That may be a very costly mistake.

Tripos offers several ligand preparation programs, including a program called ProtoPlex, for generating different protonation states and/or tautomeric states of a given compound.

“Rational drug design is a term that is evolving. The term used to be synonymous with structure-based design, but it is now adopting a much broader meaning to include other logical, calculated approacheseven ligand-based approachesto discovery.

“Many scientists have realized that blindly screening millions of compounds in the lab and hoping for a hit or a lead is an irrational drug design process that rarely pays off,” Koontz concludes.

Previous articleRace to Cut Whole Genome Sequencing Costs
Next Search for Knowledge