[gmx-users] simulated annealing

[Mark Abraham]

On 25/04/2010 10:17 PM, leila karami wrote:
> Hi gromacs users
>
> (1) Simulated annealing method is as two forms: single and periodic.
>
> I want to do Simulated annealing for obtaining global minimum in protein
> structure.

No procedure is assured of finding a global minimum with less than 
infinite effort.

> Which one of them is better? (in point of view obtaining to global minimum).
> Thank you so much for any help!

That depends on the nature of the problem. If there's a single, deep, 
unstructured energy well, then anything will work first time. Proteins 
aren't like that, typically. Several/many attempts will give you a 
better clue, but only if the starting points admit of the possibility of 
finding alternative minima. Periodic annealing might help you eventually 
find the local minimum of a very complex well, but repeated single 
annealing from a range of starting points will help you choose better 
between a set of wells. You need to make a judgement, perhaps based on 
some experience of such procedures on such protein systems.

Obviously, the reason why you want the global minimum (and of what 
quantity) will drive how hard you're prepared work in looking for it...

Perhaps you should look for some textbook and/or review material. That's 
the place to get serious discussion of this kind of issue.

Mark