It doesn’t matter if you ask a pharmaceutical company executive, an overbooked doctor or a patient seeking treatment, when it comes to discovering new drugs to treat illness, everyone agrees. We need better drugs faster.

Dr. Robert Pearlman’s drug discovery software is used by pharmaceutical companies across the world.

Photo: Marsha Miller

The drug discovery process, however, is notoriously time-consuming and expensive, requiring scientists to synthesize and test tens to hundreds of thousands—sometimes millions—of compounds in the hopes of finding one that will interact with the body in the desired way. Fortunately, advances in computer-assisted drug discovery (CADD) are improving the way scientists do their work. Dr. Robert Pearlman, Coulter R. Sublett Regents Chair in Pharmacy at The University of Texas at Austin, is a pioneer in the field.

Pearlman’s CADD software is used at almost every pharmaceutical company on the planet. One of his first programs, Concord™, has been distributed commercially since 1986, and new products continue to issue from his newly established software company Optive Research. Each software product serves a different purpose, but they are all designed to help scientists make better decisions about which compounds to synthesize and test.

“Understanding how molecules interact with one another is the key to drug discovery,” Pearlman explains. “Drug discovery boils down to finding a small molecule—the drug—that interacts in an optimal fashion with a large biomolecular ‘target’ or ‘receptor,’ often a protein. What makes drug discovery so difficult is that there are a near-infinite number of small molecules that could be synthesized and tested as potential ligands for each target.”

CADD methods assist scientists by enabling computer-based predictions of how well a small molecule might interact with a given biomolecular target even before that small molecule has been synthesized. By synthesizing and testing only those compounds that are predicted to be active, scientists can greatly reduce the number of small molecules they actually synthesize and test.

CADD predictions greatly improve the odds for drug discovery. This makes the drug development process more effective and predictable and ultimately means savings of both time and money.

Car buying is a useful analogy when considering drug discovery. If your friend were choosing a new car, how would you help her choose among all possible options? You could begin by finding out what attributes were most important to her in a car, such as body style and color.

Drug discovery begins with biology. Once doctors identify a physiological condition they would like to modify, biologists work to determine the molecular biochemistry that controls the condition. They may find an enzyme they want to inhibit, as is the case with recent arthritis research and the new class of arthritis drugs called Cox II inhibitors. Or when the objective is to cure a particular type of infection, biologists must discover a way to disrupt some process within the bacteria they want to kill without simultaneously killing the cells of our own bodies.

Once biologists discover the biomolecular “target” which controls the biochemical process they want to modify, drug discovery shifts from biology to chemistry. This isn’t surprising, once you recognize that the human body is an elaborate system of chemical checks and balances.

“Chemical reactions drive our physiological processes, our mental processes, our emotional processes,” says Pearlman. “The body is an amazing balance of chemical reactions, and it has all sorts of mechanisms for controlling the chemical equilibria which constitute a normal physiological or emotional state.

“Why do you get hungry? Because some chemical equilibrium has gotten out of balance. And you restore that balance—often just a matter of blood sugar—by ingesting some additional chemicals which we call food. Depression is due to a neurochemical imbalance. You treat it by administering drugs that affect the concentrations of those neurochemicals.”

Discovering a new drug requires taking the biological target, or receptor, that the biologists identified and discovering a compound that will fit that target and thus alter the body’s chemical state.

The needle in a haystack idiom is appropriate here. The number of drug-sized organic compounds that can be synthesized is as close to infinity as it gets, far, far beyond a trillion. Before computer-based methods were developed, scientists basically proceeded by trial and error. On average, a pharmaceutical company would synthesize 8,000 to 12,000 compounds in a single discovery effort. This could take years.

To help your friend choose what cars to test drive, you might begin by narrowing the field to only cars close to the ones your friend already likes. For example, she might like small sedans in gray, so she should begin testing small sedans in gray. Her perfect car is probably among them.

In the meantime, pharmaceutical companies were amassing samples of compounds they had synthesized and tested. In the mid 1980s, a large pharmaceutical company might have had a database of half a million compounds, represented as the familiar two-dimensional chemical structures we see in chemistry textbooks. This was useful for inventory and data tracking, but not necessarily for future drug discovery efforts. The world, as we know, is rendered in three dimensions.

“People dreamed of performing calculations on three-dimensional structures to identify those which would fit into a particular receptor,” says Pearlman. “But as late as the mid-80s, it would take anywhere from several minutes to half an hour to generate the 3D structure of just one drug-sized molecule. Faced with the daunting task of converting databases of hundreds of thousands of compounds, pharmaceutical companies never seriously considered large-scale conversion to 3D.”

Pearlman’s first commercially distributed CADD software, Concord™, changed all that. It was the first program in the world to rapidly and automatically convert two-dimensional chemical representations into three-dimensional chemical structures. The current version of the program can convert half a million drug-sized compounds into 3D structures in less than an hour. Nearly two decades after its creation, it remains the industry-standard tool for this critically important process.

Concord™ enabled exciting new strategies for computer-assisted drug discovery, the first of which was 3D-searching. After generating 3D structures of compounds in a database, 3D-searching software enables scientists to quickly search through that database looking for compounds with the right size and shape for optimal interaction with the 3D structure of a particular target. Purchasing and testing only those compounds that are most likely to interact with the target provides obvious savings of both dollars and time.

Pearlman and his team soon started thinking about how they could use the computer to generate structures of compounds that had not yet been synthesized. A number of programs followed, including a large, multi-purpose package called DiverseSolutions™.

Similar to a consumer choosing a car, scientists can find the next drug most efficiently by only testing compounds with a good chance of being viable. Compounds can be plotted on a plane using structural descriptors on each axis.

DiverseSolutions™ software helps scientists choose which compounds to synthesize from a subset of possible compounds by assigning descriptors to specific elements of chemical structures. In many cases, scientists find a number of compounds that bind to a target but don’t bind well enough to actually become drugs. These compounds are termed “leads.” DiverseSolutions™ enables scientists to use structural descriptors for these leads and plot them in a descriptor plane. The scientists can then see where those leads are clustered in the descriptor plane, and choose a subset of possible molecules to synthesize future compounds. Pharmaceutical companies responded to its introduction with great excitement and have remained enthusiastic ever since.

From the beginning, Pearlman’s goal in creating the software has been the same as that of the university: to make sure it is used.

“First and foremost, we want to see the results of the university’s research being used by the people it can really benefit,” says Neil Iscoe, who directs the university’s Office of Technology Commercialization. “Our overall goal is the distribution of technology, and the vehicle for that is commercialization.”

Since the mid-80s, Pearlman’s software has been distributed on behalf of the university by the St. Louis company Tripos Inc. Tripos pays royalties back to the university and to Pearlman’s lab.

Over time, the lab grew and started morphing into a software company of its own.

“Once you’ve got commercial software of commercial interest and you’ve got a commercial user base, you’ve got commercial expectations regarding quality and user interface issues,” Pearlman says.

Reluctant to put the pressure of such commercial expectations and round-the-clock support on graduate students and post-doctoral students, Pearlman’s lab group evolved and is now staffed entirely by full-time employees. Its ties with industry collaborators grew, creating scenarios where industry partners had unlimited access to the software in exchange for critical feedback and financial support. And after 25 years with the university, Pearlman started thinking about how to keep the software updated, distributed and developed after he eventually retired.

The obvious solution was to spin out a software company that could be self-sustaining outside of the university. Optive Research was introduced in July 2003.

Once compounds are plotted on a plane, scientists can test only those compounds similar to known leads, or compounds close to viable, as highlighted in red. This saves time and expense and makes use of existing data.

Optive’s success is good news for everyone involved. The five programs originally licensed to Tripos are now licensed to Optive but continue to be distributed by Tripos and to generate royalties to the university. The investment dollars paid by pharmaceutical industry into research at the university are returned many times over by successful commercialization.

Optive will at the same time develop new software and maintain its relationships with industry partners. Pharmaceutical companies will get the software that makes their job easier. Even the Austin community benefits.

“Most of the time drug development goes to the east or the west coast,” says Iscoe. “Optive is an example of a company that sells technology both nationally and internationally and does it from Austin. This can really help our community.”

It’s rare to find a situation in which everyone wins, but this is one. And perhaps the ultimate winner is the patient, the consumer, the individual hoping for relief from a malady or even a cure for a disease.

“The primary benefit of CADD is that you can actually develop better drugs,” says Pearlman. “CADD methodology enables scientists to devote more time to search for even better drug candidates. Rather than rushing to market with the first promising candidate, companies are now saying, ‘You know, if this compound works so well, maybe that compound would work even better.’ This leads to better drugs with fewer side effects, and it may mean drugs that you can dose once a day instead of four times a day.”

“Optive Research is not about software distribution. It’s about scientific research into improved CADD methods. We are excited about the impact our software has had and will continue to have on the drug discovery process, and we are gratified by the notion that these benefits to society are rooted in the ‘technology incubator’ which is The University of Texas at Austin.”

Vivé Griffith

Graphics generated by DiverseSolutions™ software package