Darwin’s drugs – drug discovery software – Cover Story
DEVELOPING NEW DRUGS CAN GIVE you a headache. Scientists often resort to throwing just about any biologically active substance they can at whatever malady they’re treating. Even with the most sophisticated computer analysis to help narrow down the candidates, it still takes many years to zero in on an effective treatment.
David Noever, a NASA biophysicist who helped design biology experiments for space shuttle flights, knew that the size and number of databases that contain information on biomolecules have mushroomed in recent years. In light of this, he wondered if there wasn’t some way that computers could help shorten the drug discovery process. In 1992 he stumbled across a seminar on genetic algorithms, a type of computer program that solves problems by taking a mediocre solution and evolving it through natural selection, breeding successively more fit generations of candidate solutions. Before he knew it, Noever was in the business of discovering drugs via software.
Noever’s software, which he completed in June 1997, starts off with a molecule known to provide some useful activity–say, in blocking a certain virus. Then his software produces 30 variations by randomly rearranging the atoms and substituting some atoms for others. Using the information in the database, it then evaluates for each molecule the probability that it will be effective in blocking the virus; it also “tests” for a host of criteria such as toxicity, stability, and absorbability. The highest scorers–or the “fittest”–among this generation each become a “parent” of 30 more variations. After about 10,000 generations or so–about eight days of evolution on a computer chip designed by Noever–the software spits out the structures of a handful of highly promising drugs.
Noever cofounded a company, CyberChemics in Huntsville, Alabama, to exploit the software commercially, and in June 1997 began collaborating with outside research groups that were trying to discover new drugs. The software has already led Noever to promising drugs for combating leukemia, E. coli infections, and HIV. Even if the software’s hit rate is low, it may still prove valuable. “The investment in drug research and its social consequences are so high,” he says, “that doing 5 percent better than random is a huge improvement.”
PRIVACY IN WEB SURFING
Luckman Interactive’s Anonymous Cookie
INNOVATOR: SIMON CHAN
No matter how careful you are about guarding your privacy while you cruise the Net, Web sites have a sneaky way of culling information about you: they Simply ask your Web browser to save your vital statistics–which Web sites you’ve visited, what buttons you’ve clicked on, and quite possibly your name, credit card number, and other private information–so your browser can cough it up later. Believe it or not, this feature was originally intended for convenience; rather than requiring you to type and retype this information, your browser instead holds it as chunks of data called cookies. Unfortunately, an unintended consequence of this feature is that unscrupulous Web sites can easily swipe them.
To protect the privacy of Web surfers, Marco Papa, an engineer at Lockman Interactive in Los Angeles, came up with the idea of creating a program that chucks cookies in the trash, digitally speaking. His colleague, computer scientist Simon Chan, then whipped up some software that, when activated, hunts down the files on your computer where the cookies are stored and hides them, making you completely anonymous to every site you visit. However, in the name of convenience you can deactivate the program with a click of the mouse, restoring the preexisting cookies.
Though the basic concept of the program is relatively simple, Chan says the real challenge was making it work with any version of any of the popular Web browsers, each of which stores cookies in a different fashion. To take just one of many seemingly trivial but troublesome differences, some browsers put the most recent cookies in one file and old cookies in a different file, while others keep all the cookies in the same file. Chan endowed the program with enough information about browsers currently on the market so that it could figure out automatically which browser is in use, and then cookie-hunt accordingly.
To top it off, Chan and his colleagues at Luckman Interactive let you download the program front its Web site at no charge. “We figured, why not give it away for free, since it took us only two months to develop,” says Chan. But, he quickly adds, “it was a very busy two months.”
INNOVATOR: BERNHARD GEIGER
Surgeons often make diagnoses and perform exploratory operations by slipping a flexible tube equipped with a tiny camera and headlight into the lungs, stomach, or colon. But while this procedure, called endoscopy, often sidesteps cutting open the patient, it can present problems of its own: a constricted passageway can block the tube, or a tumor can sit just out of reach of the device. In such cases, the procedure is a waste of time and there’s no way of knowing this beforehand.
Five years ago, when he was still a postdoctoral researcher at the French National Institute for Research in Computer Science and Control, computer scientist Bernhard Geiger realized that surgeons could avoid many of these wasted procedures if they could find out where tumors or other objects lay in relation to passageways. Unfortunately, most medical imaging techniques such as computerized tomography or magnetic resonance imagery, which take a series of two-dimensional “slices” of anatomy, make it hard to visualize spatial relationships. After Geiger took a job at Siemens Corporate Research in Princeton, New Jersey, he and colleagues Ali Bani-Hashemi and Arun Krishnan had an idea. If they could write software that assembled the slices into a 3-D image of the patient’s innards, they could use it to make a realistic simulation of the operation, giving the surgeon an endoscopy’s eye view of things before touching the patient.
The resulting “Fly-Through” software, unveiled last October, enables the surgeon to run through the entire operation from the comfort of a computer workstation with no risk to the patient, exploring the patient’s passageways and seeing images on the screen similar to the ones he’d get during the real procedure. The software even causes the ersatz endoscope to stop when it encounters a virtual obstruction. “The software models the actual physics of the interaction between the endoscope and the anatomy,” says Geiger.
Siemens began selling the virtual endoscopy tool through a U.S. subsidiary, Siemens Medical Systems, for roughly $25,000 a pop. In addition to helping avoid potential obstacles, it allows surgeons to plan for real procedures, and it’s a risk-free way to train surgeons. Geiger even hopes to rig up his software to endoscopy tools during a real procedure so the surgeon can track his progress on a 3-D model that maps out the planned route and target. “The software could show that the tumor is nearby even when it isn’t visible yet in the patient,” he says.
TRACKING THE CALL FOR HELP
Trueposition Wireless Location System
INNOVATOR: LOUIS STILP
Cell phones are great for reporting emergencies; 911 operators throughout the country receive more than 80,000 calls a day from the devices. But cell phone callers often don’t know exactly where they are, which makes it difficult for police or other emergency crew’s to respond quickly. A cell phone tracking system would solve the problem, but experts had long thought that the location of a phone couldn’t be pinpointed because the radio waves it emits can become muddled as they bounce off buildings, bridges, cars, and other obstacles.
Fortunately, electrical engineer Louis Stilp, a vice president at TruePosition, didn’t know anything about that when he set out in 1992 to build a cell phone locator system–initially as a child tracker, and then later for 911 calls. “We were too dumb to know it couldn’t work,” he says, so we just plunged ahead.”
Stilp and his team hit on a scheme that is essentially the opposite of using the Global Positioning System. The GPS sends out timing signals from multiple satellites at known locations to a single handheld receiver at an unknown location; the location can be calculated from the difference in arrival times between the multiple signals. By contrast, Stilps system sets up multiple receivers at known locations that compare the difference in arrival time from a single transmitter–the cell phone-at: an unknown location. To clear up the signal, Stilp’s crew came up with a mathematical formula that helps distinguish a direct signal from its reflections. That required terrific precision in determining the difference in arrival time of these signals, which, because of the relatively short distances involved, are often less than a millionth of a second apart Stilp’s team used as references the signals from global positioning satellites, which chirp at precise, predetermined intervals.
TruePosition began selling the software to telephone companies in June 1997. While it hasn’t yet been tested in the concrete thickets of Manhattan, where radio waves have myriad surfaces to bounce off, it has passed muster in Philadelphia, Houston, and Baltimore, as well as some suburbs and rural areas. “We decided we’d solve the problem for 99.5 percent of the country now, and do Manhattan later,” says Stilp.
COPYRIGHT 1998 Discover
COPYRIGHT 2000 Gale Group