A team of researchers of the International School for Advanced Studies (SISSA) of Trieste and of University of Cambridge have devised a method to reduce the time used to simulate how proteins

http://phys.org/news/2013-05-proteins-quickly.html

Quote from http://phys.org/news/2013-05-proteins-quickly.html

"We exploit the experimental data obtained observing the proteins through nuclear magnetic resonance, and use them to create restraints to be applied to the model", explained Laio, who has coordinated the research published in Proceedings of the National Academy of Sciences (PNAS).
"Basically, we used a 'trick'. Imagine I pulled your arm and directed you towards a certain place, let's say, to have a reference, from Trieste to Rimini. The trick enables to reach a destination in a period of time even a thousand times inferior to what usually required for that itinerary. Thanks to mathematical rules– gathered from previous observations carried out on 'trips to Rimini' – I can calculate, starting from the fixed travel time, how long it would have taken the same individual to spontaneously reach the same destination without being pulled. The same assumption may be applied to a protein that must fold in order to take on a certain shape."
"There is a law that connects the time a protein takes to fold using the 'trick' and the time used to do it spontaneously," explains Laio.

This reads to me like "If we know the shape of the protein, we can create restraints to make the model fold more quickly". That's ok.
But usually we fold proteins with computers because we can't obtain the shape of a given protein by different methods.
If we could obtain the shape of all proteins by measurements, the rules of their folding still would be unknown.
The real question is not their shape, but the rules make them form to their shape.
Or am I missing something here?

This reads to me like "If we know the shape of the protein, we can create restraints to make the model fold more quickly". That's ok.
But usually we fold proteins with computers because we can't obtain the shape of a given protein by different methods.
If we could obtain the shape of all proteins by measurements, the rules of their folding still would be unknown.
The real question is not their shape, but the rules make them form to their shape.
Or am I missing something here?

I doubt that their research would be published if that is all they did. More likely, the constraints obtained will allow them to calculate more quickly:

"We exploit the experimental data obtained observing the proteins through nuclear magnetic resonance, and use them to create restraints to be applied to the model", explained Laio, who has coordinated the research published in Proceedings of the National Academy of Sciences (PNAS).

I just thought it would be interesting to hear what people thought.

This is a molecular dynamics project, that predominately involves simulating molecular interaction using molecules of known structure. Basically, the structures wobble about, move around and interact with each other. It doesn't involve folding proteins of unknown structure in order to model their tertiary or quaternary structures. In fact the structures of many molecules change when they interact. The idea that a protein is a solid inflexible molecule is often wrong when it comes to interactions. For example, some surface areas of proteins interact with binding molecules in a way that guides them towards the binding site. Ditto for mediated cleavage...

---

FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

This is a molecular dynamics project, that predominately involves simulating molecular interaction using molecules of known structure. Basically, the structures wobble about, move around and interact with each other. It doesn't involve folding proteins of unknown structure in order to model their tertiary or quaternary structures.

That raises an interesting question. What does GPUGrid do? I get the impression that it is more concerned with molecular interaction rather than folding, but I don't see a clear explanation anywhere. Maybe they do both? I really don't care; whatever the good science is.

Yes, we do mostly protein-ligand interactions as you can see from our publications and a tiny bit folding of smaller proteins.
Generally, simulating protein folding at a real/large scale is at the moment still quite a dream as most proteins are quite big and fold too slow for the timescales we can simulate. If you are interested in the state of the art of protein folding simulation, D.E. Shaw's team has done probably the best ones on their Anton machine as it can simulate incredibly long single trajectories.

Yes, we do mostly protein-ligand interactions as you can see from our publications and a tiny bit folding of smaller proteins.
Generally, simulating protein folding at a real/large scale is at the moment still quite a dream as most proteins are quite big and fold too slow for the timescales we can simulate. If you are interested in the state of the art of protein folding simulation, D.E. Shaw's team has done probably the best ones on their Anton machine as it can simulate incredibly long single trajectories.

Any idea where Folding@home fits in all of this? (I'd share my GPU power to them if ONLY they'd make a BOINC-port of something of the sorts).

Folding works mostly on small to medium sized proteins trying to calculate the best fit/lowest energy shape. It's structural modeling. They also look at some molecules of known conformation to ascertain regional variation in structure, differences during an interaction. They might have diversified somewhat since I last crunched there.

It's a CPU and GPU project but you do have to use their software to fold for them.

If you want to fold using boinc try Rosetta and I would suggest anyone interested in protein folding look into or try Foldit. You can read about their research here. At present I think they are CPU only though.

---

FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Thanks for the link, tim. We are aware of this kind of research. We also are developing "tricks" to try to speed things up, but we are very careful about how we go about that. Using tricks can often mean you make sacrifices that can affect your results, which we attempt to avoid.

Folding@Home does folding of small proteins in implicit solvent, meaning that they do not have the water molecules present in the simulation. The implicit solvent model tries to approximate how water would probably interact and affect the protein. This works well most of the time, but it is known that it also has problems. The one clear reason for using implicit solvent is that you can simulate much faster.

Everything we do at GPUGRID involves explicit solvent, meaning we simulate the water, too. The water models we use are not perfect either, but are clearly superior to the implicit models, at least the ones they use now. Some people are working on improving implicit models, which, if they really are very good, would help us a lot.

I think that Folding@Home uses both explicit and implicit solvent.

gdf