Researchers using computational tools for drug design face two general challenges: one based on information, and another based on physics. The information-based challenge arises from being able to generate the information needed for a particular project, and then finding ways to handle that data. For the physics-based challenge, researchers need to determine the accuracy of the three-dimensional model and simulation software for the problem at hand. As Woody Sherman, PhD, vice president of applications science at Schrödinger in New York City, asks: “Is using a molecular mechanics force field enough? Do you need a semi-empirical method or even a full quantum mechanics approach? Can the protein be treated rigidly or is it necessary to account for flexibility?”
Beyond those fundamental issues, researchers encounter specific obstacles for particular areas of drug design. “For many small-molecule targets,” explains Sherman, “finding active compounds is not the biggest hurdle.” Instead, lead optimization might get even more benefits from computation. “Lead optimization takes significant time and money, so we could save considerably with better computational tools. It is a multi-parameter optimization. We need to find compounds that bind tightly to the target of interest while also having a desirable selectivity profile and ADMET properties,” says Sherman.
More to explore
Computational tools for drug design must accommodate the growing breadth of potential drugs. As explained by Frank Brown, PhD, chief science officer at Accelrys in San Diego, Calif.: “The obstacle today in computational approaches to drug design is that it’s not just small molecules anymore, but also proteins, siRNAs, new derivatives of peptides, and more.”
To give scientists more options, Accelrys keeps adding broader capabilities to its Discovery Studio, a suite of modeling and simulation tools. For instance, the latest version provides more technologies to work with antibodies. “We will continue to enhance these further in our next releases,” says Francisco Hernandez-Guzman, PhD, product manager for Discovery Studio. Other companies also help researchers explore more compounds. For example, Schrödinger’s Core Hopping 1.0 lets scientists take a known molecule or set of compounds and test the effect of changing the scaffold. “It’s a ligand-based tool,” says Sherman, “so you just need one existing compound.”
Beyond exploring more kinds of drugs, researchers explore new targets. “Scientists are trying to tackle more difficult problems and find new receptors to target, like GPCRs,” says Hernandez-Guzman. So Accelrys enhanced Discovery Studio with extensible homology modeling tools that can be used for GPCRs or any other membrane-protein family. Researchers can also get a glimpse of the power of modeling by making use of the DS Visualizer, which is Accelrys’s free molecular visualization tool.
Dealing with more data
Some companies started with the intent of combining data and computation.“Our main customers are in pharma, mostly large companies, but there is increasing interest from biotech and smaller pharmas,” says Jens Hoefkens, PhD, head of the Genedata Expressionist business unit for the Basel, Switzerland-based company. “We support the entire process from target identification to preclinical studies.”
The data needed for drug design come from a range of technologies, including mass spectrometry, genome sequencing, and others. “By integrating data from different sources, you can gain the insight you need for drug design,” says Hoefkens. “You also need statistical information in combination with those data. For example, for drug efficacy and toxicology, it makes sense to look at statistics on integrated data.”
Genedata Expressionist 6 helps researchers integrate data that can be used in software-based drug design. “One of the main aspects of this software for drug design is its applicability to metabolomics data, looking at various metabolite chains that come from a drug, and how these affect the toxicity of the compound,” Hoefkens explains.
Making use of metabolomics, in particular, requires large sample sizes. “You need to abstract from an individual person to a population, and that is very difficult with small samples sizes,” says Hoefkens. “You need the ability to use hundreds of thousands of samples, and our software lets you do that.”
With Genedata Analyst, researchers can perform extensive statistical tests on data. “It lets you explore data in ways you probably haven’t been able to do. It combines established statistical methods—not just t-tests, but the whole range of tools—with visualization so you can explore the data,” says Hoefkens. This software also helps researchers share data. “You can go beyond individual people doing the analysis and mining to having teams of people looking at the same data and sharing results,” he adds.
Beyond extremely specific tools, Sherman thinks that researchers need broader access to fundamental tools. For example, he says, “We need to get 3D-modeling software onto more people’s desks.” That can be done with Schrödinger’s Maestro Elements. Sherman says this was designed as a tool that “facilitates communications between modelers and chemists. It allows modelers to work in the environment where they are comfortable, and they can save annotations, scenes, and 3D molecular representations that can be passed to chemists, who can follow up with their own calculations.” Sherman adds that the “tool itself is really geared toward 3D visualization and communication with access to tasks that are most useful for medicinal chemists.”
If researchers need cheminformatics capabilities, Sherman suggests using Schrödinger’s Canvas 1.3. He says, “It allows users to build local and global models using experimental data or calculated molecular descriptors. It can handle millions of compounds if necessary.”
Other companies also focus on getting more scientists to use computational tools. For example, the Accelrys approach aims to take expert-level tools to nonexperts. Doing that requires software that is easy to use and that can work with programs from other companies. “We have lots of open source in our platform,” says Brown. “We openly embrace competitors’ software.” Moreover, many companies integrate their own code into Accelrys applications, using its informatics platform, Pipeline Pilot. No matter how someone uses drug-design software today, the computational tools must quickly handle large molecules. The next generation Discovery Studio will handle molecules composed of about 400,000 atoms. “This is the size of a ribosome,” says Hernandez-Guzman. Moreover, the software will render and manipulate the structures in real time. Perhaps most amazing of all, a customer can buy an ordinary laptop, add a graphics card, and run such applications.
A researcher looking for drug-design software can select from commercial products and shareware. “There are a number of good pieces of free software from various academic groups,” Sherman says. “The main issue is that the free software is typically hard to use. Often, only experts can use the software, and there’s typically no support.”
Sherman also says that shareware is usually less validated. “The creators may test the software on one or two systems and then publish a paper. On the commercial side, we have to make sure that a package works well and is technically robust.” To maximize the effectiveness of commercial products, they must also work with proprietary software that companies use in-house. For example Genedata’s Analyst includes application program interfaces (APIs) that let a user connect many in-house programs and make them all available from a single interface. “We also have a complete set of APIs that integrate with data sources,” Hoefkens says. “Everybody has their own LIMS, document-management system, or database, and we make it easy to connect Analyst to all of these.”
About the Author
Mike May is a publishing consultant for science and technology based in Houston, Texas.
This article was published in Drug Discovery & Development magazine: Vol. 13, No. 6, July/August, 2010, p. 8-10.