The Cutting Edge: COMPUTING / TECHNOLOGY / INNOVATION : ‘Grind and Find’ : Robots, VDTs May Be the Rx for New Pharmaceuticals
Ever since Scottish biologist Alexander Fleming stumbled on penicillin in 1929, drug development has been more serendipity than science. Researchers at large pharmaceutical companies screen thousands of compounds in trial-and-error “grind and find” tests before finally hitting on a promising drug--typically spending $400 million and 12 years in the process.
Biotechnology was supposed to change all that with breakthrough techniques such as genetic engineering. But while a few early pioneers such as Genentech and Amgen have struck it rich by “picking the low cherries on the tree,” in the words of one scientist, broader progress has been slow: Though there are about 520 biotech companies in the United States, only 26 biotech drugs are in commercial use.
Now many researchers and biotech industry watchers are pinning their hopes on a new set of drug development techniques that exploit cutting-edge computer methods and advanced robotics technology. By combining the ability of computers to analyze and model reams of data with robots’ precision handling and cutting abilities, these methods could radically reduce the amount of time and money it takes to develop new drugs--to the benefit of the sick and biotech investors alike.
“You can often make breakthroughs by marrying two separate technologies,” says Ken Lee, national director of the life sciences practice at accounting firm Ernst & Young and co-author of an annual report on biotechnology. “That’s what’s happening with high-powered computing and biotech. Something great is going to happen.”
There are no guarantees, of course: None of these technology-savvy companies have approved drugs to validate their approach, and some analysts say they lack the broad expertise in pharmaceuticals needed to produce effective drugs.
But the techno-drug companies have been producing an impressive number of drug candidates for human clinical trials, often in record time. The Johns Hopkins Oncology Center will soon begin the trial of an anti-cancer drug that BioNumerik Pharmaceutical of San Antonio developed in just 18 months using supercomputers. Agouron, a San Diego drug company, is now testing two anti-cancer drugs and a possible treatment for AIDS. Arris Pharmaceuticals, a South San Francisco-based company, is testing a drug for asthma.
And even if the current purveyors of the new approach to drug making don’t survive, few question that the technology they are developing will become part of the mainstream, accelerating drug development in companies large and small and transforming a stream of new discoveries about man’s genetic makeup into useful drugs.
*
“A hundred years in the future, everybody will be ‘genotyped’ and you will design drugs on demand fitted to that person’s target protein,” says David Galas, vice president for research at Darwin Molecular Sciences, a start-up company backed by Microsoft Corp. Chairman Bill Gates. The firm plans to combine advanced robotics with high-speed computers to sequence genes and find new ways to tackle such treatment-resistant diseases as cancer, AIDS and multiple sclerosis.
While there are many different approaches to computer-driven drug development, the broad goal is typically the same. First, the objective is to find a protein, or “receptor,” that plays a key role in a disease. Once a receptor is found, the task is then to find a compound that will attach itself to the receptor and either disable the receptor if it has harmful functions or activate it if it has a positive role. Think of the receptor as the lock that must be opened or closed for treatment and the drug as the key that must be found.
But though their goals are similar, the various companies are tackling the problem in radically different ways.
Agouron’s method, called rational drug design, is to create new compounds “atom by atom.” When Agouron set out to find a compound to disable the HIV protease, a piece of the AIDS virus that’s involved in replication, it began by cloning the protein and turning it into a crystal structure.
By bombarding the crystal with X-rays and recording the way in which the rays are deflected--a process called protein X-ray crystallography--company researchers created a three-dimensional image of the HIV protease. The image was then used to design on computer a new compound that would fit neatly into the protein, like a plug into a socket.
“We designed a molecule you couldn’t come upon in any other way,” says Peter Johnson, Agouron’s chief executive officer. “These are things that don’t occur in nature.”
A frequent criticism of rational drug design is its failure to recognize that compounds and their targets change shape under different conditions, which means the computer model sometimes bears little resemblance to reality. And X-ray crystallography doesn’t work on the many proteins that cannot easily be converted to a crystal structure.
*
Arris has tackled the problem by using pattern recognition, an approach to sorting visual information first developed by the military to help guide missiles to their targets. Arris uses the technology to find common patterns among the thousands of compounds known to make effective drugs.
For example, the system might suggest that compounds with a tower sticking up on the right, a certain electrical charge on the side and a flat, greasy surface on top will link with a particular receptor, says Michael Ross, chief technology officer. The new drugs are then designed, tested for effectiveness against computer models and further refined in continuing cycles.
“It offers completely new insights,” Ross says. “It is the breakthrough knowledge that is the most important.”
Critics say rational drug design has limited applications because it requires a huge amount of information about the molecule being studied. Such information isn’t available for most drug targets.
So another group of companies has taken an opposite tack, sometimes referred to as “irrational drug design.” They use databases of thousands or even millions of different compounds--plants, animals and synthetically produced materials--in conjunction with a range of computer techniques to separate out those compounds that have the best chance of attaching themselves to the target receptor.
Affymax Research Institute, a Palo Alto company, for example, has borrowed from semiconductor production methods to speed up the old grind-and-find approach. The company builds up layers of molecules on glass slides and covers them with a protective surface. Photolithography--a mainstay of semiconductor making--is then used to cut away areas, each cut exposing a layer of molecules to which new compounds can be selectively added.
*
In this way, a tiny chip is created that has on its surface 65,000 identifiably different but related compounds. The compounds can be tested simultaneously with a few drops of fluid containing the target receptor. Those that bind give off a fluorescent flash, which is picked up by a laser and recorded by a computer as promising compounds. The results are offered to large corporate partners for further development.
BioNumerik of San Antonio has built computer models based on fundamental tenets of pharmacology, chemistry and biology to try to predict how molecules of a given structure are likely to react. To process its lengthy equations, which are based on quantum mechanics, the company uses supercomputers and is working with computer makers to custom design computers dedicated for the purpose.
“The computers are strategic for us,” says Frederick Hausheer, BioNumerik’s founder and CEO. “We don’t have to do much testing in the lab.”
Yet another group of biotech companies are taking advantage of research coming out of the 15-year, $3-billion Human Genome Project but to tackle a different part of the equation. Instead of focusing their efforts on finding new compounds, they are looking for new drug targets.
“Over the next five to 10 years, almost all new drugs will be used against targets not yet discovered,” says Galas of Darwin, a molecular biologist who helped oversee the Human Genome Project while at the Department of Energy.
Darwin plans to develop a high-speed robotic production line that will analyze gene fragments at rates 100 to 1,000 times faster than is possible using current methods. The process yields the sequence, or code, for each fragment, and the code is then fed into a computer and matched against thousands of known proteins. If the gene’s code is similar to that of a protein known to be related to a disease, it may be a potential drug target.
The approach of genome companies like Darwin is to find new doors to a disease that might be easier to unlock.
Developing Drugs With Computers
In the past, drugs were developed using lengthy trial-and-error methods sometimes known as “grind and find.” Now computer technology is making new techniques available to speed the process.
1. Identify new drug targets:
Laboratory robots prepare and sequence fragments of genes believed to be responsible for a medical condition. Scientists then search databases for proteins--called receptors--that are similar to the genes. The hope is to find drugs that will latch on to a receptor and either disable it if its activity hurts the body or activate it if it helps.
2. Find compounds with drug potential:
High-powered computers, in combination with traditional chemistry techniques, are used to sift through thousands of compounds to winnow out those that can best attach themselves to the receptor.
3. Custom-design drug candidates:
a) Artificial intelligence is used to find common characteristics of compounds that bind to the receptor. These characteristics are combined in a single compound to develop a drug candidate better able to bind to the receptor.
b) A clone of the receptor is crystallized and bombarded with X-rays to collect information about its structure. The information is fed into a computer to generate a model of the receptor. The model is used to design a molecule that can lock onto the receptor and activate or disable it.
4. Computer modeling is used to determine how to change the molecular structure of the new drug candidate to increase potency and reduce side effects.
Sources: Arris Pharmaceuticals; Agouron Pharmaceuticals
