Re: [AMBER] graph energy check after minimization

From: Jason Swails <>
Date: Fri, 11 Jul 2014 12:19:44 -0700

On Jul 11, 2014, at 11:59 AM, amor san juan <> wrote:

> HI all,
> Would someone please give a hint on what is the possible cause of my
> strange energy plot during minimization. I attach the graphs in each 3
> minimization tests. It seems that the protein did not successfully relaxed.
> Could it be it due to amino acid clashes in the input file ? Or something
> is not proper in my input scipt.

Why do you say this is strange? Your input files all specify ncyc=500, which means the steepest descent algorithm is used to minimize for the first 500 steps (a very slowly converging algorithm). After that, the conjugate gradient method is used, which converges faster (which is why you see a sudden 'drop' about halfway through each plot).

Other than that, all I can say is that most systems are composed of at least hundreds of atoms (if not several orders of magnitude more). With this many degrees of freedom, finding a local minimum is a real challenge, so there is no reason to suspect that 1000 steps of minimization with Amber will get you to a local minimum (or close enough to make it seem like the minimization is "done").

That is rarely a problem, though, since minimization is typically used to relax bad contacts before starting dynamics -- proteins exist in room temperature, not at absolute zero, so exact local minima are not that important for most biomolecular simulations (there are some exceptions, like normal mode analyses).

If you absolutely require a local minimum, you should look into the XMIN minimizer (ntmin=3, I think -- it should be described in the manual).

Hope this helps,

Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Fri Jul 11 2014 - 12:30:04 PDT
Custom Search