So the Baker Lab out of Seattle has actually been working on that exact problem for a while now. There suit of programs for doing this type of work is called Rosetta and I know they have generated at least one protein from scratch.
They have and so has the Folding@Home team as Washington University. Although, Folding@Home is terribly inefficient at the way it approaches the problem. I know Rosetta but never have worked on it or used it so I can't comment on it's efficiency.
