Gamers PWN AIDS-Related Molecular Mystery
Sep 19, 2011 12:11 PM PT
A team of players of an online game called "Foldit" took three weeks to solve a problem in AIDS research that has puzzled scientists for years.
The problem was to solve the crystal structure of M-PMV (Mason-Pfizer monkey) retroviral protease.
The M-PMV retrovirus enables the HIV-1 virus to replicate.
The team of "Foldit" players created a model of the protease that was good enough to let researchers determine its structure. This will help in designing new anti-retroviral drugs.
How is it that gamers could solve a complex problem in a very technical discipline that had stumped highly trained researchers for years?
"I think there are two things going on here," Charles King, principal analyst at Pund-IT, suggested.
"One, 'Foldit' is a game that attracts the competitive sort of gamer who likes puzzles, and this seems to be the type of puzzle where intuition and the natural ingrained ability of the human mind to solve three-dimensional problems exceeds the ability of computers to do so," King told TechNewsWorld.
"The other is that, in many cases, scientific researchers tend to go into projects with certain preconceptions because of their training and education while gamers don't," King added.
How 'Foldit' Helped Solve the Problem
The "Foldit" players used the Critical Assessment of Techniques for Protein Structure Prediction (CASP), one of the direct manipulation tools and algorithms from the Rosetta structure prediction methodology.
Rosetta@home is a distributed computing project for creating predictions about protein structures that are close to being solved, and for designing new proteins. It's hosted on the Berkeley Open Infrastructure for Network Computing, or BOINC, run by the Baker Lab.
As of June, about 1 million members of the public had volunteered 58 teraFlops of processing power on their home PCs for the Rosetta@home project.
The "Foldit" team selected one out of five models of the protein structure produced by the Rosetta Server. After much back-end work, the team modified the model and came up with their solution.
"Whereas the Rosetta algorithm failed to notice it had produced a near-native prediction, 'Foldit' players immediately picked up on the fact that it had and correctly improved on it even further," team leader Firas Khatib, a postdoctoral researcher in biochemistry at the University of Washington, told TechNewsWorld.
The team then moved the terminal helix, some side chains and beta strands closer to their native orientation in their model of the structure.
The structure selected by the team had been ranked fourth out of five by the Rosetta server although it was fairly accurate, and the fact that the team had homed in on it despite that "is what was remarkable," Khatib pointed out.
Folding a Protein
When the primary structure of a protein turns into a functioning three-dimensional structure, that's technically known as the "folding" of a molecule.
The general process of how the molecule folds is known, but predicting the structure of the protein requires scads and scads of computing power.
"Foldit" tries to do that by getting people to use their brains' natural 3D pattern-matching abilities to make the predictions.
"Foldit" provides a series of tutorials that let users manipulate simple protein-like structures, and periodically dishes up a set of puzzles based on real proteins.
Those puzzles are graphical representations of a protein's structure and gamers can manipulate these with tools from Rosetta@home.
"Foldit" scores players based on how well their structures are folded.
More About 'Foldit'
This M-PMV crystal retrovirus structure solution is the latest achievement of "Foldit" players.
Last year at the 2010 University Protein Folding Challenge, teams of "Foldit" users competed to fold a real protein that's over-expressed in pancreatic cancer.
Corporations are becoming increasingly interested in human-based computation games such as "Foldit" because these let them leverage the wisdom of the crowd and, essentially, conduct research at minimal cost.