What is actually random?

[bruce]

EDIT: This topic was split off from here

I thought a bad WU is one where the given set of values (and their computations until the point of failure) cause the "planet to fly off the trajectory" as someone said in the old forum (bruce perhaps?).

Yes, that was a description that I wrote. It seemed to be a lot easier to visualize than atoms in a protein, but the math was similar enough for me to make my point.

Anyway, isn't this violation of limits supposed to occur at the same stage [unless caused by hardware instabilities] or is there some randomness inherent to each folding simulation of the same WU? Or are there many kinds of bad WUs? Forgive me if this is a silly question.

Well, if it's a silly question, we're both silly. I read this thread a few hours ago and decided to ask the experts the exact same question you're asking. I did get a reply. (so now I can act like I knew the answer all the time. )

All atomic motions are influenced by temperature, and heat creates atomic motions which are essentially random. (See wikipedia or any chemistry book for Brownian motion). The GROSIMT core does some internal thermal adjustments which are also based on random values and this randomness is introduced during the processing of the WU.

(If a "planet flies off into space" in one run, repeating exactly the same run may cause the planet to depart in a different direction at a different time.)

[anandhanju]

Ah. Brownian motion. Haha... reminds of my days at school when I didn't have enough material to explain this phenomenon and used a drawing full of squiggly lines

Doesn't the random nature of the simulation introduce a possibility of something (an incorrect fold perhaps, due to a random progression of variables) occuring in real life that does not occur/cannot be predicted in x simulations? Darn, I cannot phrase that correctly, will edit this later if no one understands a word of it.

I reckon this is enough deviation from the OT to warrant an EUE

[uncle_fungus]

Ah. Brownian motion. Haha... reminds of my days at school when I didn't have enough material to explain this phenomenon and used a drawing full of squiggly lines

Doesn't the random nature of the simulation introduce a possibility of something (an incorrect fold perhaps, due to a random progression of variables) occuring in real life that does not occur/cannot be predicted in x simulations? Darn, I cannot phrase that correctly, will edit this later if no one understands a word of it.

I reckon this is enough deviation from the OT to warrant an EUE

The key is statistics. Individual WU trajectories on their own are meaningless, the trajectories only become meaningful when analysed statistically with all the other thousands for that particular project.

[bruce]

In vivo, the key is also statistics.

If you've ever watched the PS3 graphics, you'll understand that the protein spends a large amount of time NOT folding. The motions are essentially random -- until all of the statistics just happen to all line up pointing the same way. Then ALL OF A SUDDEN the protein will drop into an energy hole and take up a new shape that is one step closer to being folded.

Read some of the early papers on how and why FAH works and how the process can be paralleled rather than depending on a single serial trajectory.

I'm going to split this topic off and put it in the science section so that the original topic can continue without distraction.

[anandhanju]

Thank you, bruce and u_f. I think I understand what you said to a certain extent. Is this the same thing? http://www.unb.br/ib/cel/chico/artigos/ ... ode18.html (REM not to be confused with the SREM core; I had an AHA!... oops... moment there)

...Then ALL OF A SUDDEN the protein will drop into an energy hole and take up a new shape that is one step closer to being folded.

That does sound very dramatic. Is there a PS3 video grab of such an event? I, unfortunately, do not have a PS3 and the vids I saw online have the atoms behaving like bumblebees after a coffee break (states of higher entropy, I assume; random motion of the atoms not leading to anything).

Alrighty, back to arguing about PPD now...

[VijayPande]

Check out the movie in the middle of
http://www.youtube.com/watch?v=EZ1XuOgknuE
(go past the DL TV stuff). I tried to explain what folding looks like in that trajectory. I use one of my favorite analogies ("folding is like a drunk guy trying to parallel park"). Drunks play a large role in Brownian motion analogies, but definitely do not try the drunk parallel parking at home!

[anandhanju]

Thanks Vijay. I'd actually chanced upon that clip earlier but didn't get past the DLTV stuff It was interesting watching the protein reach the final configuration for that fold. What prevents it from unfolding right away and attempting to refold? (To use your drunk guy analogy, what makes him realize that he's parked correctly and doesn't need to attempt it again?) The bonds: do they act like child-safe locks?