A Map That Maps Gene Functions
By Kristen Philipkoski
2:00 a.m. May 28, 2002 PDT
The genetics revolution is generating such a gigantic glut of information that artificial intelligence may be the only way scientists will ever put it to practical use.
Inspired by an AI effort to record all of the common-sense knowledge shared among humans called Cyc, scientists have come up with a technology that can gather all of the information scientists know about an organism.
This body of information, like the amount of common sense humans retain, is inconceivably huge. Therefore, researchers are starting small, with tiny organisms like bacteria.
By learning everything there is to know about a bacterium, scientists can learn how to kill it efficiently, or change it into, say, a pollution-eating bacterium.
"Because of the huge amount of information being gathered by biologists, it's impossible for one person to grasp everything that's known about an organism," said Peter Karp, director of the bioinformatics research group in the Artificial Intelligence Center at SRI International.
"So it's important to bring together all the information known about one organism in a form that computers can manipulate it and make it accessible to scientists," Karp said.
Karp's conception of how to do this is a platform of database software called Pathway Tools. The software gathers and maps out the biochemical pathways used by genes in an organism to keep it alive.
Working with Doug Lenat -- who started the Cyc common-sense project in 1985 -- inspired Karp to apply some of the Cyc artificial intelligence techniques; namely, using knowledge representation to map metabolic pathways in organisms.
Although some consider the Cyc project to be disappointing, Karp has adopted some of its artificial intelligence concepts to create software that many scientists are beginning to rely on.
Karp started out with the E. coli bacteria in 1992. He and his colleagues had to enter all of the information manually.
But now that it's in there, Pathway Tools can "computationally infer" pathways to help build other databases. SRI is building them for tuberculosis, H. pylori -- the bacteria that was recently discovered to cause ulcers -- and 10 others.
"It may well have taken years had we had to follow the same manual approach as we used for EcoCyc," Karp said.
Instead, it took just weeks.
The database can be viewed as a physical map that shows the metabolic pathways as lines and the metabolites (products or key biochemicals of metabolism) that connect them as geometric shapes.
"Each line represents a function -- drug companies want to ask 'which line can we cut to kill the organism?'" he said.
The geometric shapes around the periphery are "transporters" that take small molecules from the outside to the inside of a cell.
Gray lines represent pathways that haven't yet been identified, and could lead to the discovery of new genes. In come cases, companies are interested in these because they can be pretty sure they're not patented.
In E. coli, probably about 30 percent of the metabolic pathways are not yet identified, Karp said.
The Human Genome Project has often been called a map. In reality, what the researchers have identified are like cities, but without the roads that connect them. That's an endeavor they may never finish. And because the list of cities alone consists of about as many letters as 2,000 copies of War and Peace, a full map would take up several buildings.
The "pathways" generated by Pathway Tools indicate the routes the biochemicals in an organism take to metabolize food. They lead to gene function. Cutting off just one of these pathways could kill the entire organism.
Or, changing one or more of the pathways could lead to an organism that, instead of infecting humans, eats harmful carbon dioxide in the atmosphere. The Department of Energy has given Karp's group funding for such work.
"In order to re-engineer a bacteria, you have to know what you're starting with," Karp said.
Researchers at the Carnegie Institution in Palo Alto, Ca, are using Pathway Tools to map the genome of a small plant, a member of the mustard family called Arabidopsis.
Analyzing its pathways could eventually lead to better herbicides or hardier plants.
"Artificial intelligence comes in when you can use the tool interactively and ask very advanced queries. Others just implement perl scripts that are, I'd say, dumb," said Lukas Mueller, a researcher at the Carnegie Institution and a curator of the Arabidopsis Information Resource.
Other databases that attempt to identify the metabolic pathways in organisms include KEGG from Kyoto University in Japan, and WIT, which stands for "What Is There," from the Argonne National Lab.
Karp hopes Pathway Tools will become the standard for looking at genetic pathways, and researchers say it's heading in that direction.
"Understanding pathways is critical to understanding the biology of an organism, and if Pathway Tools continues to develop and improve, it may be widely adopted," said Steven Salzberg, senior director of bioinformatics at The Institute for Genomic Research.
"We'd like every important organism to have databases like this," Karp said. "What's the point if you don't keep looking for those functions that are not yet determined? We have to keep updating it for the scientific community."
Copyright
© 1994-2002 Wired Digital Inc. All rights reserved.
See also:
• Sea Squirt Savants Celebrate • Genetic Fate Is in Venter's Hands • DNA Plus Bacteria Equals Art • Read more Technology news • Check yourself into Med-Tech
Related Wired Links:
Genome Map on a Grain of Rice
March 29, 2002
An 'Atlas' to Count the Genes
Feb. 27, 2002
Expert: Look at the Clean Genes
Feb. 26, 2002
How to Turn On a Gene
Feb. 6, 2002