HCC-Small: Protein Design Through Massively Distributed Video Games

University Of Washington, Seattle WA

Investigators

Abstract

In the past few decades biochemists have discovered that practically all fundamental molecular design problems are about the geometric relationship of complex molecules. Massively parallel and distributed versions of such algorithms have been recently developed in hope of finding such proteins with shear brute force. For example, implementing screensavers that harvest idle CPU time from PC users worldwide to provide sufficient computing power to solve the protein prediction problems. So far, all these attempts have produced very modest improvements. This project takes a radically different approach: the molecular folding problem is cast as a massively distributed 3D puzzle game, and encourages people to work together with computers to find the solutions to current open problems including cures for cancer, AIDS, and discovery of novel biofuels. The fundamental idea is to employ user-assisted optimization for protein design, and formulate and present it as a competitive game played by thousands of people. The intention in this project is that people will play the game beacause it is fun (it looks like a fun 3d puzzle and not like some biochemistry textbook), it is addicting, it is competitive, it is collaborative (players can work together in groups to solve a problem), has impact (players want to get credit for the drug that cures cancer). The impact of this proposal is in its addressing of a number of fundamental areas: 1) how to best develop games to maximize human ability to discover novel proteins beyond what is currently possible with computation-only approaches; 2) determining the guiding principles of a successful molecular design game; 3) how to best generalize game-development principles to the widest possible range of biochemical problems; 4) revealing what is learned from the way people play the game, and how these strategies could be "distilled" towards developing stronger automated approaches.