Simbios Talk by Adrien Treuille, Carnegie Mellon University, October 21, 2009

Title: Scientific Discovery through Computer Games

One of the most important discoveries of biochemistry in the last two decades is that many molecular chemistry problems are effectively problems of geometry, or of how different shapes fit together. The folding of proteins and RNA, for example, are both fundamentally 3D puzzle problems. However, the conformational search space is so large that despite the use of massive distributed networks, solutions remain constrained by global computational resources.

This talk presents a completely new approach to solving these important problems. Our fundamental insight is that although computers are very good at exhaustive searches in small subspaces, they are not effective when the search space is very large. Conversely, humans are very adept at visualizing 3D space and solving very large visual problems by using abstractions and macro manipulations in space. Therefore, the optimal algorithm for protein design would harness both the power of human 3D abstract reasoning and computer's exhaustive search in small subspaces.

To this end, we have been developing a suite of computer games to harness human creativity for scientific discovery. Already, over 90,000 registered players now compete on-line to find low energy protein structures. We are currently developing follow-up games based on protein design, small molecule drug design, and RNA evolution. Looking further, we hope these games inspire a new generation of scientists to apply large-scale human-computer cooperation to our most pressing scientific problems.