Folding Forum

I have seen occasional references (in some of these threads) to trajectories. Apparently there are two types of trajectories - "long" trajectories and "short" trajectories. I read something on (Wikipedia?) about the Anton supercomputer being good at doing "short" trajectories while FAH excels at doing "long" trajectories - or maybe it was vice versa. Intuitively, I think a generic trajectory might be one (of many) possible ways a protein molecule folds - although such a description might make a trained microbiologist cringe.

I would be grateful if somebody could explain (in a not overly technical way) what "trajectories" are and how trajectories tie in with FAH simulation. (An explanation of the difference between "short" trajectories and "long" trajectories would be greatly appreciated.)

Thanks,

Alan

Some of what you're looking for may be found here.

bollix47 wrote:Some of what you're looking for may be found here.

I think a better and more up-to-date reference is: http://folding.stanford.edu/home/faq/faq-simulation/
https://en.wikipedia.org/wiki/Folding@h ... gnificance, and the pictures in that section, are also useful.
Neither of those answer the question directly.

F@h is good at long trajectories - simulations on the order of milliseconds. Anton can do long simulations like that too, but one thing that F@h does well is statistically detailed simulations. For example, someone could fire up Anton and run NTL9 for a few milliseconds and see a single folding trajectory. Alternatively, they could fire up F@h and see a multitude of possibilities of many, many possible folding trajectories that all arrive at a common state (or not). These could include the one that Anton saw, but many more besides. One could then look at the F@h Markov model, find the Anton results, and see that Anton's trajectory had a 70% chance of happening, but the remaining 30% consist of many other paths that are less likely to occur. There are many ways for a protein to fold, and molecular forces, temperature, or a tiny nudge could cause one path to happen instead of another. Having the bigger picture is very handy.

Actually we usually say that F@H produces many, short trajectories! You are correct that we use markov modeling to "stitch together" the short fah trajectories to study long timescale dynamics

To answer OP, "trajectories" are all the positions of atoms over time. It is essentially a "movie" of the protein. Long vs. short means exactly what you'd expect in this context

Picture is worth a thousand words: https://commons.wikimedia.org/wiki/File ... @home.tiff

Each of the lines between states is a trajectory. Note that this is a simplified version of the actual protein, which in reality had thousands of states and trajectories between them.

This is a trajectory of a moderately complex protein showing actual atoms. You'll notice a lot of seemingly random motions but if you compare the beginning state with the ending state, you'll see some rather significant changes.

This is an animation of the folding trajectory of a simpler protein where several common structures are symbolically represented. It's a lot easier to see the progress.