The Next Top Protein Model

Web Extras

Article: Teamwork Opens Evolutionary Window

Article: Now Open for Drug Chemistry

Article: Protein Puzzlers

From building muscles to healing wounds, our bodies rely on proteins—chains of small molecules called amino acids that fold into unique shapes. Incorrectly folded proteins can cause disorders like sickle cell disease or cystic fibrosis. Ever-improving computer power is making it easier for researchers to predict how proteins fold and interact with other molecules, possibly leading to new treatments for protein-related disorders.

»  Tailor-Made Proteins
»  Modeling@Home
»  Project Structure

Tailor-Made Proteins
By Emily Carlson

Scientists can easily determine a protein's amino acid sequence, but they can't reliably predict how this sequence will fold into a three-dimensional structure.

Baker used his computer program to design a small protein not found in nature.
Credit: Brian Kuhlman, Gautam Dantas, David Baker

So computational biologist David Baker at the University of Washington in Seattle took a different approach. He started by sketching a protein structure that nobody had ever seen. Next, he relied on a computer modeling program he developed called Rosetta to tell him what amino acid sequence would form the three-dimensional shape of his made-up molecule. Baker used that sequence to build an actual protein that was stable and quite similar in structure to the one he had drawn, validating his approach.

With the ability to whip up new proteins, Baker's research may make it possible to customize proteins that could be used as drugs or tiny biological machines to treat certain diseases.

Back to Top

Modeling@Home

In high school, Johnathon Tinsley had mixed feelings about math and science. "Math was very challenging," he recalls. "I enjoyed some parts of biology, but not physics."

While you're sleeping, your computer could be doing scientific research.

Today, this British teenager is helping to find cures for diseases like AIDS and Alzheimer's just by letting researchers use his computer when he isn't. You can get involved, too!

Tinsley is part of a tech trend called distributed computing that relies on the public to help advance health and medicine. Through this approach, researchers harness the power of personal computers to answer important questions about biology. The typical computers in a scientist's lab can't perform all of the required number crunching, but a network of hundreds and even thousands of personal computers can.

How It Works

You join a distributed computing network by downloading free software. When your computer isn't busy, it sends a message to a server in the researcher's lab basically saying, "Hey, I'm available. Can I help?" The server assigns a chunk of a large calculation that it knows the home computer can solve. The donated computer may spend several days working out the problem. When it's done, it hands in the answer. Just like teachers, people in the lab check the result, also making sure that no one has tampered with the information.

You can volunteer your computer, whatever the make or model. The computer must be connected to the Internet—the type of connection doesn't matter. Older computers can do the job, although they generally get simpler calculations. You can also choose how much computer memory you want to donate.

You don't need to worry about hackers breaking into your computer system. Security checks protecting the main servers and the limited capabilities of the required software make participating in the projects considerably safer than surfing the Internet.

Before you download distributed computing software onto a public computer, like the ones at school or work, ask if it's OK. If you don't, you could get into serious trouble!

Distributed Computing in Action

"The science we can do is unmatched by what we could do with any other available tools," says Vijay Pande, a scientist at Stanford University in California who started a distributed computing project called Folding@Home.

Pande studies the dynamics of how proteins fold into their unique shapes. By studying how they fold, Pande can see what goes wrong and how drugs might patch misfolded proteins.

Proteins fold much faster than you can fold a shirt. The quickest one is done in just 5 millionths of a second.

Pande says that it would take a very fast desktop computer more than a thousand years to completely simulate the process! But with the help of nearly 200,000 personal computers participating in his project, Pande can do the job in about a week.

Tinsley donates about 40 hours of processing time every day between his two computers. Tinsley likes knowing that his computers are doing something useful. He says, "They're not just sitting there like stuffed lemons"—British slang for being idle.

For his distributed computing projects, Tinsley tracks how much work his computer has contributed compared to others'. If his computer helps predict a protein structure, he'll see his name on the project's Web site and maybe even published in a scientific journal. Some projects also award special certificates.

"Seeing the impact makes a big difference," says Pande. "When you donate to many charities, you don't see a direct link between what you give and how it's used. For us, you can actually see what your computer has donated and the results."

Serving science, though, is not the only benefit. Distributing computing also offers its participants an active social network. Many projects have message boards where donors can post questions about the science or random thoughts about life.

Donors who really want to be ranked at the top often will form competitive teams.

"I like competing to get my stats above my team members'," says Tinsley. But he also really likes the social aspect. For one team, he explains, "The main aim is to meet and talk with friends and do something good and worthwhile while we're at it."—EC

Back to Top

Project Structure

The computer model generated by David Baker's team for the 2004 community-wide experiment (left) was strikingly similar to the protein's actual structure (right).
Credit: Philip Bradley, David Baker

Most people enjoy a little friendly competition, and protein structure prediction researchers are no exception. Every other year, these experts go head-to-head to see whose computer models make the best predictions.

The goal is to most accurately model the shapes of pre-selected proteins. The contestants don't know the actual structures of these molecules, but the judges do. After reviewing the entries, the judges invite the most successful modelers to an international meeting where they talk about the approaches they used. The entire group discusses how all can do an even better job in the future.

The scientists don't actually call the event a "contest" or even a "competition." It's a "community-wide experiment" to improve the accuracy of protein prediction modeling so researchers can discover new drugs more quickly and cheaply.—EC

Back to Top