Proteins are trying to get into their most “comfortable” position; that is, where they are at the best energy equilibrium with their environment. Many interactions drive what shape a protein will adopt. For example, some proteins contain areas that are hydrophobic (hate water), so those sections of the protein will end up away from the aqueous environment by hiding in the middle of the folded protein. There are many other factors that drive the protein, but there are several different analogies that can be used to explain the general process.
For one, think of a huge beach ball bouncing down the side of a steep mountain. The ball will bounce many times as it descends and will eventually stop. If you throw the beach ball down it again, there will be random variations in its path and it won’t end up at the same place. If you repeat that process many, many times, you can determine that there is a statistical pattern to the final resting points. You can also see a statistical scattering in the amount of time it takes for the ball to stop. Most of the time the ball will end up at the bottom of the mountain, but occasionally it may end up in another nearby depression and never reach the lowest possible stopping point. There is a significant statistical nature to atomic motions in proteins, much like the motion of the ball bouncing down the mountain. Normal folding is represented by all the times that the ball ends up at the lowest point. Misfolding is when the ball ends up somewhere else.
In some respects, it’s also similar to parallel-parking a car in a crowded street. At first, the car is exposed, and it usually takes several steps to park the car in the proper position. Sometimes it may be necessary to back out slightly and then try again. A protein does the same thing. If an observer watched a hundred similar cars being parked in that space, they would come to understand the common ways of parking and which methods work and which don’t.
Like both examples, it’s important for us to know about the motion of a folding protein, although we also want to know the intermediate steps along the way. Our simulation methods construct models of both of these properties. One aspect that makes Folding@home different from some other distributed computing projects (Rosetta@home, for example) is that we want to see how the car parks, not just the end state of seeing it parked. While that’s an important result, it doesn’t shed any light on how or why a protein sometimes misfolds. By attempting to study all of the possible paths that the bouncing ball can take down the mountain, we learn a lot about the question “How did we get here?” It also gives us the opportunity to introduce changes – such as drugs – into the process that modify the probability of misfolded results.