A scratch on the old vinyl record? Repeatative loops seen.

[Scott Jensen]

I notice when bored and checking on my text client via the window that ever so often (rather frequently) the program seems to get caught in a loop. It keeps repeating two numbers (say 32/76 and 33/76) over and over and over. It is a bit amusing to see how far these repetitions go. Once they scrolled all over the screen and colorized practically all of it. Then all of a sudden it seems to click and speed through the rest of the numbers. This loop work has taken sometimes up to a minute to break out of. Is this desireable? Is this a problem that will eventually be weeded out of the client program? I assume if it is an undesired result, that eliminating it would speed up search. In the time many of these loops occur (a couple times lasting up to a minute), over a hundred other simulations/searches/whatever-you-call-them could have easily been done. Or, again, are these loops what's desired in the first place?

[ColinT]

I've always thought this was a good thing. Don't know why though...

[Aegion]

Originally posted by Scott Jensen
I notice when bored and checking on my text client via the window that ever so often (rather frequently) the program seems to get caught in a loop. It keeps repeating two numbers (say 32/76 and 33/76) over and over and over. It is a bit amusing to see how far these repetitions go. Once they scrolled all over the screen and colorized practically all of it. Then all of a sudden it seems to click and speed through the rest of the numbers. This loop work has taken sometimes up to a minute to break out of. Is this desireable? Is this a problem that will eventually be weeded out of the client program? I assume if it is an undesired result, that eliminating it would speed up search. In the time many of these loops occur (a couple times lasting up to a minute), over a hundred other simulations/searches/whatever-you-call-them could have easily been done. Or, again, are these loops what's desired in the first place?

To the best of my knowledge, the numbers are simply an estimated amount of progress on that particular structure, alhough they do reflect the amount of ammino acids inside the structure. What happens is that typically the algothem is able to detect that that particular guess of what the folded protein looks like is definately off, and therefore it terminates that guess and starts a new one. What is usually happening in the cases where it repeats two numbers is that the guess is actually close and the program is attempting to get as detailed a guess as possible in order to ensure that the structure closely resembles the actual shape of the folded protein. If the numbers repeat, this is actually good and means there is a good chance that the RMSD of that guess should be quite low.

[Brian the Fist]

Ive mentioned this before but cant hurt to say it again (sometimes I feel like that protein, repeating myself over and over :p )
This is because the algorithm "backtracks". This i easier to see on the graphical screensaver. When it tries to place a residue and atoms collide with existing atoms, it backtracks and tries a different placement. If it gets itself into a really tight spot (generally indicating a compact, protein-like structure - a good thing) it can get 'stuck' for a while until it finally backtracks enough to get out of the tight spot and worm its way in a another direction. This behaviour is perfectly normal, and intentional, and assures us that we do not get structures which tie themselves in knots, or are physically impossible because two pieces of it overlap each other in space.

[Scott Jensen]

[QUOTE]Originally posted by Brian the Fist
[B]Ive mentioned this before but cant hurt to say it again (sometimes I feel like that protein, repeating myself over and over :p )

Which usually indicates a good question to add to the FAQs. ;)

[Nofinger]

Originally posted by Scott Jensen
[QUOTE]Originally posted by Brian the Fist
[B]Ive mentioned this before but cant hurt to say it again (sometimes I feel like that protein, repeating myself over and over :p )

Which usually indicates a good question to add to the FAQs. ;)

It does can be found on the DF homepage only not in the FAq but here (http://www.distributedfolding.org/details.html)

"Why does the protein seem to get 'stuck' sometimes?

The protein is built using what is called a 'kinetic random walk'. This means that as the protein atoms are added one atom at a time, if a collision between two atoms occurs, it will backtrack and try a different placement. This avoids building unrealistic structures that cut through themselves or tie themselves in knots. Generally the more compact and globular a structure is, the more collisions it will have while building and the higher the chance it may get stuck for a while. It will almost always complete eventually though. Since real proteins are compact and globular, when it gets stuck this is usually a good thing as it means your structure is probably compact and protein-like, although of course it could be a completely incorrect fold. It gets hideously stuck and cannot get out, the program will time out and start a new structure after about 8 N log N seconds, where N is the number of residues in the structures - this works out to about 15-30 minutes depending on N. "

[Scott Jensen]

Nofinger,

I had read "stuck" as not moving as all and not repeating a section over and over.

[Nofinger]

Originally posted by Scott Jensen
Nofinger,

I had read "stuck" as not moving as all and not repeating a section over and over.

I was searching for a anwser for a question about this subject and read over it also but after a good look I saw that it was a answer for that question. :sleepy: ;)

What I wanted to say with this text above is, I know what you mean :D
it can be phrased beter and be placed in the FAQ to make it easier to be found.