[gmx-users] simulated annealing

[Mark Abraham]

On 21/04/2010 10:19 PM, leila karami wrote:
> Dear Justin
> thanks for your accuracy.
> I want to use simulated annealing to obtain global minimum for
> protein-dna structure. I will do first heating and then slowly cooling
> and I will my full md simulation in 300K. is my manner true?
> if so, is my mdp file true, now?

Try it and see. Note that a single annealing procedure is not guaranteed 
to find the global minimum - no non-exhaustive algorithm can. I'm 
struggling to see why being there would matter for a starting structure 
for MD.

> Tcouple = berendsen

That still won't work, as Justin told you.

> ref_t = 300
> annealing = single
> annealing_npoints = 9
> annealing_time = 0  3  6  9  12 15  18  21  24
> annealing_temp = 150  650  600  550  500  450  400  350  300
>
> my main questions are that
> 1) is last line of mdp file suitable for simulated annealing?

If you'd done your homework and read some literature where other people 
had done such annealing processes, you'd have a good idea whether this 
is true.

> 2) is there relation between (annealing_time = 0  3  6  9  12 15  18
> 21  24) and (dt and nsteps)?

Yes. 7.3.16 of the manual clearly shows what that is.

Mark