why the protein folding problem isn't really a problem anymore

You may have read the occasional article in the New York Times science section, or BBC online, about the 'Protein Folding problem' and how it's one of the last great frontiers of theoretical physics. That might have been true 15 years ago but, although no biophysicist will admit it in public, the problem has been solved for some time now. I'll explain:

Proteins bend and twist into a unique three-dimensional shape which determines their properties and utility in metabolism and cellular structure. Which particular three dimensional shape a particular protein takes is usually determined solely by it's amino acid sequence. What that means is that the forces which cause a protein to adapt it's conformation are the result of a combination of different kinds of intramolecular forces, i.e. hydrogen bonding, electrostatic interactions, van der waals interactions, and the hydrophobic effect.

These forces are at this this time well enough understood to be modeled accurately on a computer. By this I mean that small organic molecules like phenol can have their structures transcribed into a program like GAUSSIAN 99 (a quantum mechanical program which calculates wavefunctions of molecules or sets of molecules) and the program will predict not only their conformation but the strength and conformation of their interactions with other molecules.

These kinds of calculations are extremely computationally intensive, and the computational cost increases exponentially with the number of atoms in the system of molecules which one models. Obviously then a full protein, which contains upwards of 1000 atoms in a favorable case, is not accessible to today's computers. This is why for the last 30 years or so many biophysicists have been trying with little success to find broadly applicable simplifications of these quantum mechanical models so that proteins can be modeled successfully. However, despite some unfortunate claims, this has still not worked.

During this time period computational power has increased exponentially (see Moore's law.) Small peptides can now be modelled accurately just as simple things like phenol were 10 years ago without any simplifications or approximations. It doesn't take a genius to decide that within a few years, entire protein structures will be modeled quantum mechanically.

IBM has figured this out, and it wants to get there first. Which is why it's building blue gene. When they start determining the structures of all these new genes that we don't know anything about that have appeared lately due to the human genome project, they will patent them immediately, and make bucketloads of pharmaceutical money on the new drugs designed on the basis of these structures.