Canadian researcher turns the tables on illness with software to seek out drugs' hidden abilities
Aug 21, 2007 04:30 AM
The cure for an emerging outbreak may already be in your medicine cabinet.
And with an artificial intelligence computer program he's creating, a Canadian researcher is honing in on that link.
"By the time (an outbreak) occurs it's already too late to create a new drug from the ground up," says University of British Columbia chemist Artem Cherkasov.
"The idea is when a new infection comes in or gets noticed, to use some of the existing substances and try them against the virus or pathogen."
Cherkasov presented his program, now in the later stages of development, at an American Chemical Society meeting on Sunday in Boston.
He says there are untold numbers of drugs on the market that may have more than one pharmaceutical function. His program aims to identify those additional medical capabilities and match them up against emerging infectious ailments.
"My program predicts in a virtual world ... what the chances are for all those tens of thousands of substances to be antibiotic," he says.
"So before testing anything, I can prioritize and rank all those compounds and once they are ranked you have a list of higher priority entities to test against a new pathogen."
The pharmaceutical sales titan Viagra, Cherkasov says, proves his point.
"It was created as a cardio condition drug ... and it had a useful side effect people noticed that made it famous," he says of the hugely popular erectile dysfunction medication.
Cherkasov has used his software to pour through huge compendiums of known drugs and has successfully come up with substances that have unheralded antibiotic properties.
But he's not yet tested his computer against such potential outbreak viruses as avian flu. He says he hopes to have it up as a functioning system up by 2009.
Drugs that have already been sanctioned for human use, unlike new medications, can usually get rapid approval for new purposes from regulatory agencies like Health Canada or the U.S. Food and Drug Administration, Cherkasov says.
The program technology he's employing is known by the general name "in silico." It is increasingly being used by drug companies to test compounds against diseases in a virtual computer world.
"Traditionally the way drug design worked, people would use a trial-and-error approach where they would test millions of chemicals to find suitable ones in a lab," he says.
Cherkasov, who has no connection to any drug manufacturer, says his software has been programmed to "learn" which substances may look like antimicrobial drugs.
Testing a set of drugs with known antibiotic properties against a set of other substances, Cherkasov has "trained" his program to distinguish, with 95 per cent accuracy, which agents might fight infectious ailments and which will not.
"And once you're happy with your training accuracy, then you can just pour in unknown compounds," says Cherkasov, an assistant professor in UBC's infectious disease department. "And your pre-trained artificial intelligence system will assign probabilities for the unknown compound to be either antibiotic or not."
In separate research, funded by Genome Canada, Cherkasov is also working on technology that can "open up" pathogenic agents with the use of mass spectrometry and find likely sites on the germs or viruses where antimicrobial drugs might readily bind.
"Once the pathogen-specific targets are identified inside the organism, then we'll use this in silico approach to come up with a first line of defence in a computational lab," he says.