Probably better stated as "Why the protein folding problem may some day be a thing of the past." The protein folding problem has two major components which overhauser mentioned. One is the issue of accurately representing all the forces that cause a protein to collapse and arrange in the right conformation. The other is a combinatorial problem. Our best methods of folding a protein currently entail exploring the vast multidimensional conformational space using Monte Carlo techniques, genetic algorithms, simulated annealing, threading or any of a myriad of other computational techniques.

Using a quantum mechanically (QM) derived forcefield may solve the first problem, giving an accurate representation of the forces that glue the protein together. However, it does not necessarily address the combinatorial issue, which is a purely computational problem. With current approximate molecular mechanics forcefields, it is already a daunting task to simulate folding of anything but the smallest protein domains. If a QM forcefield were introduced instead, the time required to simulate folding would be enormous.

If Overhauser's utopian view of the future of protein folding comes true, and both forcefield and computational power reach a state where we can ab initio determine the final structure of a protein based on its sequence, then the sky is the limit. With an exact forcefield, we can have a full understanding of dynamic and functional properties of a protein. The current approach to protein folding issues and structural genomics is based more on a knowledge-based paradigm, rather than a theoretical one. I believe that the time of proteins folding under exact QM forcefields is a long long way off.

Of course, I hope I'm wrong.