AMBER Archive (2006)

Subject: AMBER: Not getting proper structure after minimization

From: Dave, Sonya (
Date: Tue Oct 24 2006 - 15:50:04 CDT


I am trying to make a structure of a protein. I have the crystal structure of a related protein, and the sequence of my protein. I had InsightII calculate a homology comparision. I am now trying to minimize that protein, to get the final structure.

We are generally very confident this protein contains a motif called the CBS domain. This is, in the following order, beta sheet - alpha helix - beta sheet -beta sheet -alpha helix. PFam agrees it has this motif. However, when I minimize, I can not get this order to show up. I am only modeling the CBS domain of the protein, so that domain is basically all of the structure. I am very tempted to believe I am doing something wrong modeling, rather then that the protein does not contain the domain.

Any ideas of how I can improve or change my model? I know this is a VERY broad question, but I have very little experience modeling, and would appreciate suggestions.

Here is what I am doing.
I always use a cutoff of 12 um.

1. Minimize molecule in vacuum (steepest descent 500 cycles, 100K cycles with steepest descent, followed by conjugate. done per tutorial)
2. Put molecule in octahedral box of size 25um.
3. Hold protein fixed, and minimize water.
4. Minimzation of 500 cycles, with water fixed (per tutorial).
5. Minimization with whole system.
6. Molecular dynamics with weak restraints on protein.
7. Molecular dynamics without restraints (100Kcycles)

I get a structure with linker regions, and 2 alpha helixes followed by 2 beta helixes.
Basically, I am following the tutorial, except minimizing in vacuum before putting in a water box.

The final product shows potential energy stabilized between -276000 and -277000
The RMSd from the original structure is stable, around 2 A.
The movie of the last molecular dynamics shows the atoms are still moving around at the last stage. Is this typical?
The density stabilizes around 1.08 - 1.1 mg/ml

To me, it looks, by all criteria a good model (Except for the part of the movie). So, do any of my parameter look improper, or is there any way I can improve this model (besides run more cycles). General ideas, thought, etc welcome.

Thank you all for your help.


The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to