RE: AMBER: some questions about minimization and MD of DNA duplex with sander

From: Ross Walker <>
Date: Wed, 31 Jan 2007 08:36:27 -0800

Dear Wendy
1. Minimiztion steps:
Before the MD simulation, 7000 steps (2000 for minimizing water box, 5000
for whole system) of minimization is applied. The RMS value of last step is
1.5584E-01. Is it enough? Should I further minimize the system when RMS is
less than 1.0E-03, or even 1.0E-04?
This is reasonable for starting MD from. The only reason for doing
minimization before MD is to remove any bad contacts that exist in the
experimental structure and also clashes from adding protons and water
molecules. The main aim is just to remove the largest forces so you should
be good here. If the Md runs fine then don't worry about it. Obviously if
you wanted to do normal mode analysis then you would need a much better
As an aside you may find it more useful to look at the GMAX value rather
than the just the RMS force value when minimizing since this shows you the
largest force. Compare the GMAX value on step 1 and step 5000 of your
minimization and you should find that it has dropped a lot.
2. Analysis of md result:
I noticed that average structure of the overall simulation may not be
reasonable for comparison. Can I take the conformation with lowest energy as
"simulated structure"?
Some published papers provided a analysing method: use CARNAL module of
amber to determine average structures for different conformational families.
But I can't find "carnal" at AMBER9, not even the manual mention this. Is
"CARNAL" avaliable in AMBER9? If not, is there any other module can be used
Average structures are really not very useful in themselves. E.g. consider a
rotating methyl group the average position of those atoms is with the
protons all on top of each other along the axis of rotation. Using the
lowest energy simulated structure is perfectly reasonable. You might also
want to consider carrying out cluster analysis on the system. This will
allow you to essentially bin structures into different families. This can
often be quite useful as for example it might show you that to get from one
structure to another the system always has to go through the same
intermediate structure first. It will also give you an idea of the lifetime
of each structure class. It can be fairly subjective due to how big the
clusters are that you pick etc but I think that using it as a guide as to
what to look for is reasonable. An example of how to do this is in Tutorial
4. minimization of average strucuture:
Even I get the average structures for different conformational families, the
minimization of average strucures is still needed. So the question of
minimization steps still exits. I reviewed the reply of similar questions at
AMBER maillist, most recommend a small number of minimiztion steps. But I am
confused. Why not minimize the system to meet converge critia, like RMS=
1.0E-03, or 1.0E-04?
It depends on what you want to look at. If you are just interested in the
structure then minimising so much probably won't gain you anything.
Similarly at 10-3 RMS force the energy is probably converged to something
like 6 decimal places so do really need the energy that good? If you want to
try it yourself try minimising to an RMS of 10^-1. Save the structure and
energy and then minimize further until you get to 10^-4 - then compare the
energies and structures. Now if you interested in second derivatives then
the argument is very different.
I hope this helps.
All the best

|\oss Walker

| HPC Consultant and Staff Scientist |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- |
| <> | PGP Key
available on request |

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Sun Feb 04 2007 - 06:07:11 PST
Custom Search