You are here: Home » News » Biotechnology » Online gamers help UW scientists tackle complex protein problems

Online gamers help UW scientists tackle complex protein problems

August 6, 2010 | By Michael McCarthy More

Online gamers are helping scientists tackle protein-folding problems that can even stump super computers, University of Washington researchers report in this week’s issue of the scientific journal Nature.

Seth Cooper, a doctoral student in computer science at the University of Washington, was the paper’s lead author. Zoran Popovic, associate professor computer science and engineering, led the project.

The project, an online game called Foldit, is one of a growing number of research programs that are enlisting the help of online “citizen scientists” to solve tough scientific problems.

In this case, the problem was predicting what shape a protein will take after it has been synthesized.

Proteins are made of simple chain of amino acids, and today it is relatively easy to determine the sequence of those amino acids.

But before proteins can do their job, these chains must fold into very precise shapes, and figuring out what that shape is can be very hard, indeed.

An amino acid chain folds into a functioning protein spontaneously. Credit: Dr. Kjaergaard Wikipedia

That’s because how a chain folds is determined by an incredibly complex set of interactions between the individual atoms within different parts of the chain as well as atoms in the surrounding fluids.

In some cases, the interactions draw atoms together, in others the atoms repel each other, driving them apart.

In the end, the sum of these forces drive the protein chain to spontaneously fold into its stable, functioning shape.

When the protein chain has assumed that shape, it is said to be at its lowest energy level.

Screenshot of the UW protein-folding game Foldit

Now, scientists know the general rules guiding the folding process.

In general, sections of the protein chain that are hydrophobic (water-fearing … think “oil and water”) fold into the center of the protein to be out of contact with the fluid surrounding the protein, while those sections that are hydrophilic (water-loving) tend to stay on the outside, where they will be in contact the surrounding fluids.

And there are other factors: the amount a segment of the chain can bend or twist may be limited, for example, and bonds can form within the protein linking sections together

So even if you know the sequence of amino acids in a protein chain, because so many interactions involved in the folding process it is still extraordinarily difficult to predict a proteins final shape.

One strategy to solve this problem is to use computer modeling programs that try to predict how a particular protein chain is likely to fold to achieve its lowest energy level, which is likely to result in the shape it would assume in nature.

The problem is that to take into account all the interactions involved in folding, the computer must perform trillions of calculations, which can take a very long time even for the fastest supercomputers.

(And for researchers getting time–and paying for time–on these expensive machines can be a challenge it itself.)

One increasingly popular strategy to get around this problem is to farm out the number-crunching calculation work to home and office computers, which can perform the calculations when they are idle.

That’s how the Foldit project got started.

One of the paper’s authors, David Baker, a UW professor of biochemistry, had developed a protein-modeling program called Rosetta.

Baker’s team took that program and created a version, called Rosetta@home, that anyone could download and run on their home and office machine.

While the program crunched the numbers, it created a model of the protein it was working on as a screensaver so people could see what their computers were doing.

Some 200,000 people downloaded Rosetta@home, but soon some were asking for a way to manipulated the proteins themselves because they could see the program was making obvious mistakes.

So teaming up with Professor Zoran Popovic’s team at UW they set out to create an online game to see if humans could help computers solve these protein-folding problems, and Foldit was born.

The approach was inspired in part by the multi-player online game “World of Warcraft” in which thousands of gamers from around the world work together to, well, wage war on each other.

In Foldit, improperly folded protein structures are posted online. Players, working either alone or in teams, can then interactively reshape the structures in directions they believe will lead to the lowest energy lower.

The lower your structure’s energy level, the higher you or your team scores in the game.

Foldit presents simplified versions of the protein, highlighting “energetically frustrated where the player can likely improve the structure,” the researchers write in the Nature article.

Players can manipulate protein with a variety of maneuvers, including such maneuvers as ”tweak”, “shake” and “wiggle”.

More than 57,000 people have played the game, many of whom have little or no background in biochemistry.

In the Nature article, the researchers describe an experiment they devised to see just how good humans were at this game.

In the experiment, the researchers posted a series of 10 puzzles and then looked to see who got the best scores, the humans or the Rosetta program.

Click to view a Nature video on Foldit

The humans did surprisingly well, scoring higher scores for five of the 10 puzzles and doing just as well as Rosetta in three other puzzles.

Humans were particularly good when the solution required major remodeling of the protein’s backbone, which was sometimes required to make it possible take an exposed hydrophobic structure and tuck into the center of the protein.

Humans were also more likely than the computers to take a protein through maneuvers that substantially raised the protein’s energy level, gambling that in the end the maneuvers would ultimately lead to a lower energy conformation.

The researchers also found that the human players would use much more varied exploration methods than computers and will alter their strategies as the game progresses and the challenges of the task change.

In teams, humans tended to specialize the researchers write, “Within teams, there is often a division of labor; some players specialize in early stage openings, others in middle and end game polishing,” they said.

The Foldit experience gives an idea of what is possible with such “hybrid human-computer” efforts, the researchers conclude, and demonstrates what might be achieved “if even a small fraction of the energy that goes into playing computers games can be channeled into scientific discovery.”

To learn more:

Read the Nature paper (subscription or payment may be required).

Watch the Nature video on YouTube.

Go to the Foldit website where you can learn how to play.

Read “People Power“, a news article in the August 4 issue of Nature in which Eric Hand describes the growing number of scientific research projects that are enlisting the help of “citizen scientists” to tackle tough scientific problems.