QSAR has been around for a while, but is making a comeback in a form that is better than ever.

click to enlarge Fragment-Based Drug Discovery (FBDD) as well as ADMET must consider how adjacent polar fragments may interact. Metalaxyl illustrates how the hydrophilic character of an ether, amide, and ester is reduced by proximity and the aromatic amide made more polar by steric hindrance: ether-amide +1.33, and ester-amide +1.21, and ‘steric twisting’ -0.80 to give CLOGP = 1.85; Meas. log P = 1.65. (Source: Albert Leo, PhD)

Prediction can be fun. People are enamored with those who can predict the future; and, that’s probably how psychics stay in business. But, in the world of science—hard science—the word oracle is surely met with skepticism. Indeed, in the physical and life sciences, predictions have to be based on experimental evidence. As a life science, pharmaceutical science is also dependent on experimental evidence for prediction. But, pharmaceutical scientists have built their own oracle—it is called QSAR (for quantitative structure-activity relationship). This predictive approach was invented in 1963, when Corwin Hansch and his colleagues at Pomona College (Claremont, Calif.), studying Hammett-Taft methodology using select try parameters to better understand reactions in solution, used a hydrophobic parameter to provide the missing link for applying Hammett-Taft methodology to organic biological reactions. Since first being published in 1964, the work became an instant classic, and publications containing the work have been cited thousands of times in scientific literature.

Hansch’s QSAR method, which eventually became known as 2D QSAR, was the single best method for using linear and nonlinear equations to predict the effect of the hydrophobicity of a compound on its transport through the body. “From the 1960s to the 1980s, QSAR was a crucial method for trying to design new biologically-active drugs using the Hammett-Taft method,” says Albert Leo, PhD, president and chief executive officer of BioByte, a developer of QSAR methods based in Claremont, Calif. In the 1990s, however, as high throughput screening and combinatorial synthesis took hold, QSAR briefly fell out of fashion. But it quickly made a comeback: “The pendulum has swung back that way, and fragment-based drug discovery is coming back into its own now using a wider range of physicochemical properties than was provided by the Hammett-Taft methodology,” says Leo. And, now, high throughput screening and QSAR are being used together.

Leading by QSAR
QSAR methods are used to perform lead optimization of small-molecule pharmaceuticals, which is the final step of the drug discovery process. “After putting in a lot of work, time, effort, and money into developing a lead compound, you hope that you can identify a compound that you can take into clinical trials,” says Richard Cramer, PhD, chief scientific officer, Tripos, St. Louis, Mo. “What is needed is a good way to rank the compounds found through lead optimization. So we are interested in searching for R-groups within databases that people are already using for docking compounds.” Cramer is the inventor of the original 3D QSAR, which is called CoMFA. He explains that 2D QSAR and 3D QSAR differ in the way their descriptors are generated, and in the statistical methods they use. For example, the calculated structural properties that are entered into a column of 3D QSAR table as potential explanatory descriptors will be different than those entered into a 2D QSAR table.

Cell-QSAR? Stefan Balaz, PhD, professor of pharmaceutics and biopharmaceutics at North Dakota State University in Fargo, is also a developer of QSAR methods. His research does not focus on any particular compound or disease; rather, he is trying to develop QSAR techniques that will combine the well established, ligand-based or receptor-based methods for isolated receptors with a description of drug transport, to generate the conceptual cell-QSAR models for cell-level data. “My lab developed the disposition function, which is a nonlinear function of physicochemical properties, such as solvation in the headgroup and core regions in the plasma membrane bilayer; acidity; hydrolysis rates; and structure describing structure-specific binding to transporters and enzymes; to describe intracellular disposition of drugs within cells,” says Balaz. “This function describes the intracellular disposition of a drug depending on its dose, physicochemical properties, and exposure time.” Besides developing equations for QSAR methods, Balaz is also using these methods to understand the physical biochemistry of flexible, conserved binding sites in metalloproteinases and other metalloproteins, some of which have pharmaceutical and medical relevance.

And, as a scientist working for Tripos, Cramer has developed the latest flavor of 3D QSAR called Topomer CoMFA. With Topomer CoMFA, it is possible to perform virtual R-group screening, which—within the QSAR approach— enables drug researchers to seek compounds with improved biological activities by searching among the candidate R-groups embedded within structures of whole molecules. And, because the name of the game with QSAR is to produce new chemistries by altering the structure of R-groups, Topomer CoMFA has been designed to meet that purpose. “One of the goals of virtual screening is to find new chemistries that have the same or improved biological effects,” says Cramer. “And then we are using the Topomer CoMFA approach to provide the best rankings of compounds in a very cost-effective manner.” QSAR is also able to perform these more accurate rankings within minutes, not weeks, as would be required using most other computer-aided drug design rankings, which is how QSAR saves time and money for pharmaceutical companies.

click to enlarge (Top) 3D-QSAR model of CYP2C19 explaining why (here, the right-hand side of) acetohexamide should not be active on this antitarget. Green and yellow blobs indicate regions where steric changes apparently increase and decrease potency, respectively. Red and blue blobs indicate regions that favor or disfavor negative charge (or disfavor or favor positive charge), respectively. Arrows indicate the particular field dependencies that are determining the activities of the ligands shown. (Source: Tripos) (Bottom)3D-QSAR model of Estrogen receptor activator indicating the structure-activity relationships of the active Sulfamethaxole versus an inactive neighbor compound. (Source: Tripos)

QSAR also enables the drug researcher to determine the physicochemical requirements for inhibition of a particular target. That’s how it is being used by Robert Doerksen, PhD, assistant professor, University of Mississippi, Oxford. One particular protein target that Doerksen has focused on is glycogen synthase kinase 3 (GSK3). Doerksen’s research hypothesis is based on previously published data that have shown that a series of malamide compounds have strong inhibitory activity against GSK3. Doerksen mined these data and used QSAR (both 3D and 2D QSAR) to study the patterns in physicochemical requirements for binding of malamides to GSK3. He found good structure-activity relationships for the different malamide compounds. One of the reasons that Doerksen studies GSK3 is that it is a potential drug target for malaria and several central nervous system disorders in humans. “The malamide work is just a starting point for further development of novel lead structures against GSK3,” says Doerksen.

Doerksen believes that statistical analysis is a large part of using the QSAR approach. “When people report a series of 80 to 100 compounds, they look for activity, but they don’t go to the next step of performing any formal statistical analysis,” he says. Doerksen uses several commercially-available software packages for statistical analysis, including SYSTAT (Systat Software Inc., San Jose, Calif.); Cerius2 and Catalyst (Accelrys, Inc., San Diego, Calif.); and sybyl (Tripos). Doerksen has successfully used both 2D QSAR and 3D QSAR to analyze ligand-target interactions, and is likely to continue using these approaches in the future.

Understanding ADMET
Obviously, one must have a lead compound before using QSAR to develop it. QSAR enables the drug researcher to understand how the hydrophobicity of a lead compound will affect its absorption and distribution, i.e. its ADMET properties—and that’s a good thing—as often these properties can mean the difference between success and failure of a clinical candidate.

And, of course, the wisdom today is that the earlier a compound can be removed from development, the better, because it saves time and money. By using QSAR, a company can make that prediction in advance of spending all that time and money, making it a very powerful tool for drug development.

About the Author
James Netterwald is president and CEO of BioPharmaComm LLC, a provider of writing, editing, and consulting services to the life science, pharma-biotech, and public relations industries.

This article was published in Drug Discovery & Development magazine: Vol. 12, No. 5, May, 2009, pp. 26-28.