IBM's goal is to build the Blue Gene computer at the petaflop scale in just a few years, and to apply it to solve the ProteinFolding? problem. The project will focus on modeling the folding process of a protein of average size: about 300 amino acids. The number of possible three-dimensional shapes for a protein is greater than the number of atoms in the universe, and the folding model will encompass more than a billion forces acting over a trillion different time steps. Blue Gene will be 500 times faster than the fastest supercomputer today, but it is still expected to take around one year of continuous, around-the-clock processing to solve the folding problem.
Uses similar technology to DeepBlue but might become obsolete by the time it is finished in 2006 by others InSearchOfClusters or GridComputing.
This is a lot of effort. It might it be more efficient to table the protein-folding problem until QuantumComputing can have a go at it?
Folding can't be too hard. We living things have been doing it for a long time. What the difficulty really shows is the smallness of our theory of computation.
Through history, living things have caused amino acids to be folded in different ways basically by chance. It's only over millions of years that structures inside cells arose that could fold them mostly the same way every time - by creating a suitable environment around the acids conducive to certain folds. Living things have never folded proteins consciously for a particular goal: that will be an achievement made by our computers.
Could not the process be used the other way - tap into the complexity of ProteinFolding? somehow and use it to compute? Gene transcription has apparently been used to do certain algorithms like the TravellingSalespersonProblem?.
We already know that the answer will be "42".
Currently the world's fastest SuperComputer - http://www.llnl.gov/PAO/news/news_releases/2005/NR-05-06-12.html
See also GeneticCode