Friendster for Proteins

Friendster for Proteins
Robert Langreth and Matthew Herper 03.12.07

Understanding how the body's tiny components communicate is opening up vast territory in drug research.

Peter Sorger spent eight years developing new laboratory gadgets and arcane mathematical theorems to explain how networks of genes and proteins can go awry, causing cancer, arthritis and other diseases. But when he went looking for cash to start a company in 2000, the best year ever for biotech venture capital, the vcs wouldn't give him decent terms.

Triumphant drug researchers assumed they had everything they needed. The human genome had just been mapped, and they had the supercomputers and tools to figure out which genes caused diseases. Who needs to bother with understanding how one gene interacts with thousands of others, let alone parsing the relationships among the hundreds of thousands of proteins these genes produce?

But seven years later drug approvals are down, not up--and Sorger's firm is suddenly hot. The body-as-network is becoming a dominant metaphor for future drug research. Sorger, a Harvard biologist, is one of the leading lights in a budding field called systems biology, which borrows techniques from engineering, physics and computer science to understand the body's complex web of cause and effect. The company he cofounded, Merrimack Pharmaceuticals in Cambridge, Mass., has raised $140 million from nontraditional sources including medical device billionaire James Sorenson and has an arthritis drug in human trials.

Theories about protein networks have been overlooked for decades, maybe because the idea is so obvious. Every gradeschooler knows the hip bone connects to the thigh bone. But scientist-radicals such as Sorger say the biotech industry's dogged pursuit of isolated genes and proteins offers few clues to the function of the whole.

"Can you imagine trying to predict how a computer chip is going to work when you are missing 90% of the circuitry? Yet that is exactly what the pharmaceutical industry does when it tries to make drugs," says Colin Hill, a physicist who heads Gene Network Sciences in Cambridge, Mass.

The body contains 25,000 genes, which can form as many as 1 million proteins interacting in hundreds of millions of ways. Perhaps only 5% to 10% of all protein and gene interactions have been documented so far. Network biologists' ultimate aim, still a decade away, is to create computer programs that could simulate the effects of drugs on cells in the same way that Boeing (nyse: BA - news - people ) simulates a new jetliner before it flies.

"It is the dark ages of the drug industry, and we have to make it more like building cars or planes," says Trey Ideker, a bioengineer at University of California, San Diego, who is working on a computer simulation of a budding cancer cell. Adds Sorger: "When something is big and complicated you cannot do it all by anecdote. No single person understands every last wire in a 747."

Leroy Hood, who invented automated gene sequencing, pioneered the network approach and set up the Institute for Systems Biology in 2000 in Seattle. In 2003 Harvard Medical School deemed the field worthy enough to make systems biology its first new department in two decades.

Big Pharma is starting to pay attention. Pfizer (nyse: PFE - news - people ) aims to use the new approach to develop drugs that hit more than one node in the network at once. Novartis (nyse: NVS - news - people ) has 50 researchers trying to simulate the molecular networks that cause cancer, Alzheimer's and diabetes. Merck (nyse: MRK - news - people ) has committed hundreds of scientists to its network effort; one early result led Merck to spend millions to license a promising cancer drug from Vertex Pharmaceuticals (nasdaq: VRTX - news - people ). A half-dozen biotechs tout their ability to map bio-networks on their computers.

Biologists have long known that diseases such as cancer and diabetes are caused by a network of faulty proteins gradually going awry. But the complexity of these networks is beyond what anyone had expected. A recent analysis found 122 mutated genes involved in breast cancer and 69 in colon cancer. Scientists had estimated far fewer.

Notre Dame physicist Albert-László Barabási and others have shown that molecular networks inside our bodies share some basic similarities with the Internet. Typical proteins make relatively few links to other proteins, just as most Web sites link to only a few others. But a small handful of "hub" proteins are like the Google (nasdaq: GOOG - news - people ) of the protein world, connected to dozens of other proteins. "Disease is a breakdown of the network," he says.

This big-picture idea was posited in the first half of the century by mathematicians and engineers. But the notion was eclipsed by the 1953 discovery of the structure of DNA. Since then molecular biologists have adopted an atomistic focus, rushing to identify and understand one gene and protein at a time.

It is only within the past ten years that a new array of high-speed genetic tools has made network maps possible, by allowing researchers to look at thousands of genes or proteins at once and measuring how they interact. Then researchers can go back to cells or animals to test whether the predictions hold true.

Concrete results are sparse, and systems biologists worry there is too much hype. "Anybody that thought the genome was going to directly provide drugs was a fool," says Leroy Hood. "Biological networks are not simple, and making drugs to affect them won't be simple. Drug companies don't really understand how far away we are." Quips Novartis research head Mark Fishman: "Some of my best friends run systems biology departments. They still haven't been able to explain to me what it means."

A few substantive ideas have emerged. Entelos, of Foster City, Calif., is helping drug companies simulate clinical trials for diabetes, asthma and other diseases using virtual patients that exist only inside computers. A two-year trial can be run in a few hours. Johnson & Johnson (nyse: JNJ - news - people ) says the Entelos program correctly predicted that a new class of diabetes drugs it was testing in mice would be ineffective in humans. "That sort of thing makes you start to be a believer," says Michael Jackson, who heads J&J's West Coast research labs.

Subscribe to Forbes and Save. Click Here.