Some of us are involved in the folding@home project, spending time, money, and resources.
I would like to know the answer to two main questions:
in silico modelling of anything in biology is an active field of research. It's very useful for making predictions and developing hypotheses, but any findings need to be confirmed experimentally.
From the Folding@Home website:
Folding@home has been a success. In 2000-2001, we folded several small and fast folding proteins with experimental validation of our method. We are now working to further develop our method, and to apply it to more complex and interesting proteins and protein folding and misfolding questions. Since then (2002-2006), Folding@home has studied more complex proteins, reporting on the folding of many proteins on the microsecond timescale, including BBA5, the villin headpiece, Trp Cage, among others. In 2007, we crossed the millisecond milestone by simulating a protein called NTL9, and the 10 millisecond barrier in 2010 with ACBP.
More recently (2006-present), we have been putting a great deal of effort into studying proteins relevant for diseases, such as Alzheimer’s and Hunntington’s Disease. You can learn more about our results and peer-reviewed scientific achievements on our Papers Page.
As they say, you should check out their research contributions for how it's being practically applied.
You may also be interest in this article for the science behind and prospects of protein folding prediction: Bowman GR, Voelz VA, Pande VS. 2011. Taming the complexity of protein folding. Curr Opin Struct Biol 21(1):4-11. At least try reading the conclusion for a less technical overview.
I will try answer your question directly.
How do we know if we fold it right?
A. If you're interested in only the end product pf folding -- the 3D structure, then this is the subset of the folding problem called the structure prediction (from sequence alone).
a. We can verify the structure experimentally by determining the 3D structures by NMR or crystallography. In fact, the CASP competition withholds newly solved structures and see which algorithm comes out the closest to the experimental structure.
b. Based on Anfinsen's hypothesis, we can calculate the free energy of folding, and the structure that has the lowest energy among the assembles is probably (but not necessarily) the folded state.
B. If you're also interested in how exactly the protein folds, including its folding rate, plausible intermediates etc.:
a. You can experimentally determine the folding rate although the computed folding rate is often off by several order of magnitude.
b. Some interactions, esp non-native interactions (aka. interactions between amino acids that are only observed during folding but not in the final folded structure) can be verified by carefully designed NMR experiment.
1b. Yes it is possible. Adjusting the force constants and force field in molecular dynamics (MD) (the core engine of Folding@Home) so that the calculations converge to consistent and sensible results is not trivial.
2b. If you mean the end of product of computation model, then probably no, but one example came close. A model determined purely computationally (aka. ab initio), although still not as useful as an experimental structure, was close enough to facilitate the experimental determination of that structure (by a method called molecular replacement in X-ray crystallography)
A certain fold of a peptide string can be validated or ruled out if other experimental data is available. Some other techniques to infer protein structure are X-ray crystallography (requires pure protein that will crystallize) and single particle analysis.
