February 15, 2006 (Vol. 26, No. 4)
Visualizing the Interactions Between a Potential Drug and Its Proposed Target
LigandScout, released this month by Inte:Ligand (www.inteligand. com), features a 3-D pharmacophore modeling approach that, according to Inte:Ligand, offers significant time savings when compared to ligand docking. A pharmacophore search for millions of compounds can be performed within hours, notes Gerhard Wolber, Ph.D., CTO, while it takes several weeks to dock these compounds into one single target.
Inte:Ligands family of software products accelerates drug development at early stages and reduces the number of potentially failing candidates in later phases, where failure may become really expensive, says Dr. Wolber.
Time savings is not the only benefit to in silico modeling. For example, At Argenta Discovery (www.argentadiscovery.com), we have found that in silico design helps provide more cost- and resource-efficient drug discovery, according to David Clark, Ph.D., director, CADD and knowledge management. Other advantages to in silico design are the ability to reduce the number of wet experiments needed, he says, as well as the possibility of replacing some animal experiments with suitable in silico models.
But, all computational models are based on assumptions and approximations that are compounded by a frequent lack of consistent and accurate experimental data upon which to base a model,adds Dr. Clark. In structure-aided design, there are still key classes of targets for which it is not yet possible to obtain structures routinely. For example, although some 3040% of drugs act at targets in the G-protein-coupled receptor family, there is currently no experimental structure available of a therapeutically relevant GPCR that can be used to guide drug design.
Despite such challenges, in silico design returns great value by allowing researchers to focus on library subsets that contain all or most of the hits available from the full library. As Chris Burns, Ph.D., research director, Cytopia Limited (www.cytopia.com.au), notes, a 100,000-member compound library with a 0.1% hit rate can be reduced to a 1,000-member subset and screened, identifying 90 of 100 possible hits and achieving a 100-fold cost savings.
Structure-based design is helping drug development companies visualize the interaction between a potential drug and its proposed target with the aim of refining a promising molecule to enhance its characteristics, thus improving efficacy, specificity, or pharmacokinetic properties, asserts Dr. Clark.
The ability to understand and rationalize observed structure-activity relationships in terms of the fundamental molecular recognition that is occurring between the ligand and the protein allows researchers to propose alterations to the ligand that may improve its potency or pharmacokinetic or physiochemical properties without sacrificing potency.
Now that x-ray structures of important drug metabolizing enzymes are becoming available, it is beginning to be possible to consider rationally designing out metabolic liabilities from potential drug compounds, Dr. Clark says. Virtual screening is another viable option, in which large collections of molecular structures are computationally docked into the appropriate binding site and then ranked according to how well they fit and thus, how likely they are to show activity against that target.
Using LigandScout, pharmacophores for small organic molecules can be derived within 0.1 to 30 seconds. Protein-peptide and protein-protein interaction take a bit longer. LigandScout offers advanced PDB ligand perception and manual corrections while modeling in the active site, intuitive pharmacophore-based molecule alignment based on pharmacophoric points rather than atomic contributions, unlimited undo levels, and 2-D depiction linked to a 3-D editor.
3-D pharmacophore screening enrichment is at least comparable to classical docking methods, Dr. Wolber says. The relevant interactions are transparent to the researcher and can be reviewed and modified if required. Because LigandScout is interoperable, results can be exported to major software packages. Inte:Ligand is developing an activity profiler for specific pharmacological groups to screen potentially problematic compounds, as well as an improved screening platform.
The biosciences group of Fujitsu Computer Systems (www.us.fujitsu. com/biosciences) has developed a new, multifaceted approach to ADMET modeling that predicts the metabolites of a particular compound and assesses the probable degree of risk associated with that metabolites production. It combines structural activity modeling with a knowledge base approach that predicts likely compound metabolites by the CYP450 enzymes and then combines that data with predicted interaction with varying CYPs via docking, according to Ian Welsford, Ph.D., manager, application science.
Dr. Welsford says this system avoids the constraints and inaccuracies associated with mathematical modeling and data-based approaches. We use QSAR-based approaches with a curated knowledge base approach coupled to a CYP-focused docking and modeling approach. This results in a systems biology-based appreciation for the tuple relationships found among the CYP proteins. Therefore, it can handle both derivatives of existing scaffolds, as well as make high quality predictions about new scaffolds. We validated our method using actual animal test data on a family of compounds obtained from a commercial collaborator.
The lack of any real system physiology understanding among many of the other approaches was partially behind the development of this new method, according to Dr. Welsford. Fujitsus approach allows researchers to make predictions that account for quantity of interaction between the compound and the CYP superfamily. We are now including interactions with other entities, such as drug efflux pumps, and adding this into the mix to further refine our modeling capabilities in this area.
Cytopia introduced the ChemaPhore suite of software to help research potent, selective inhibitors based on virtual screening of 3-D proteins structures from x-ray, NMR and homology modeling, binding mode analysis, and SAR generated through medicinal chemistry and wet screening. Each software module is specialized for only a few applications, like data preparation docking, scoring, data analysis and comparison. All modules can understand the data and the results the others produce, notes Herbert Treutlein, Ph.D., head of the computational biology group.
ChemaPhore can dock one molecule at a time, but its full power comes into play for virtual screening of compound libraries. When we analyze data, we include not only scores and results from a single one-docking run, but look at all results and data available, possibly including experimental results for all compounds in a library. Scoring includes detailed information about the specific binding site of a target, and, to some extent, its flexibility, adds Dr. Treutlein.
We are not really focused on predicting exact binding affinities of single molecules, but on the question of how to make the most out of the available databoth virtual and realto help enhance the compounds, Dr. Burns says. Cytopias synthetic chemistry is focused on high-value compounds. When comparing efficiency at predicting hits with ChemaPhore versus actual wet-screening data, efficiencies have ranged from two- to more than ten-fold, depending on the diversity of the screen set examined.
De Novo Pharmaceuticals (www.denovopharma.com) is removing some of the inherent uncertainties by incorporating protein flexibility and chemical tractability enhancements in Reflex, the new incarnation of SkelGen. Designed for work in-house and with collaborative partners, Reflex incorporates flexibility, selectivity, and improved assessments of ligands, thus allowing a schema to be produced to determine how the compound can be synthesized. In the kinase family, for example, Reflex has helped develop specific ligands for each known ligand binding site and for promiscuous ligands throughout the entire family, says Nikolay Todorov, Ph.D., principal scientist.
So, if you have a related protein, you can design a compound that is selective for one and not the others, adds Philip Dean, Ph.D., CSO. As you build up a small compound structure in a site, you allow the site to flex on the fly, unlike previous rigid models, says Dr. Dean. Reflex handles induced fit designs, which allows these models to change throughout their development.
Competitive Workflow, by Cyprotex (www.cyprotex.com), takes a different approach to in silico design by modeling and automating what people actually do, according to David Leahy, Ph.D., consultant and former CTO for Cyprotex.
This software architecture provides exhaustive modeling by implementing every step, multiple times and reacting to changes. New data or methods cause the whole process to reactivate. When new ADME data is added, for example, the whole process is rerun automatically, says Dr. Leahy. Modeling is tightly coupled with our experimental data so is specific to chemotype. It captures the expertise of human scientists in software and is a learning environment, so successful methods are reinforced. We think we can extend that application to most in silico design processes.