Computing Trajectory

Moderators: Site Moderators, FAHC Science Team

Post Reply
theteofscuba
Posts: 96
Joined: Wed Dec 05, 2007 7:15 am
Hardware configuration: PS3, Phenom II X4, QX9775, HD 8570
Contact:

Computing Trajectory

Post by theteofscuba »

How many times in a typical WU does the software try to pull a random number out of a hat? Is the randomness done in the hope that the calculation "wins the lotto" so to speak? If randomness is frequent, how much more data can be processed if it were to simply iterate through increments instead of randomizing the trajectory path?
bruce
Posts: 20910
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: Computing Trajectory

Post by bruce »

The answer is not the same for every FahCore, but there's a much deeper question underlying the concept that you're asking about.

At any temperature above absolute zero, thermal energy causes motions that are best characterized as random. A real protein has to "win the lotto" to fold, but lotto tickets are reissued so frequently that it can accumulate enough wins very quickly to reach the desired folded shape. In a solution, multiple molecules won't all take the same path from unfolded to folded.

Any non-randomized trajectory does not represent the actual physical process, it's not "the solution" that we're looking for. That's a huge reason why the folding process is not well-understood. To understand the overall process, you have to understand the statistics associated with all possible changes of shape from unfolded to folded. That's why FAH's research needs such a huge are such a huge job.

There's an area of mathematics that deals with how to create mathematical models which accurately represent random processes. That's one part of answering your original question, and I don't think you really want to get into that.

In simple terms, finding a single trajectory that happens to pick all the right lotto tickets will fold very rapidly but it would be the "wrong" answer, scientifically. First, it would be much too fast than would ever be seen in real life. (FAH's results have to agree with experimental data.) Moreover, to understand how a few percent of a protein can misfold you obviously don't want to just know just one idealistic trajectory.

The statistics generated from a few hundred or perhaps many thousands of trajectories create an accurate picture of a very complex process, and it's the patterns that can be seen in those statistics that reveal new insights into the overall process.
Post Reply