RE: AMBER: QM-MM output interpretation

From: Ross Walker <>
Date: Mon, 30 Jul 2007 08:51:36 -0700

Hi Francesco,

> settings as in Tutorial A2 (may be a comma is lacking in
>; more

You are correct, well spotted, thanks. I will fix it on the website.

> The structures (mdcrd) are OK, except for poor performance in
> dealing with the
> conformation of highly conjugated (non aromatic) cyclic
> systems, as expected
> for PM3 and all other semiempiricals.

You may want to try DFTB here (qmtheory=7) this may do better for these
types of systems although I have never tested it for this.

> I am unsure how to read the outputs. As to relative
> populations of the two
> conformers,

I'm not sure what you mean by relative populations. A single snapshot from
the MD does not tell you anything about population, it simply gives you the
energy (kinetic and potential) of that particular snapshot. Also note that
the SCF energy in itself is not that meaningful in a periodic boundary
simulation where you have PME turned on since the PME field perturbs the SCF
energy, due to the inclusion of the PME potential in the fock matrix, as
does all the MM atoms in the direct space sum. Then part of the energy due
to the PME field is essentially 'recovered' in the Ewald terms and so the
SCF energy on it's own is not complete. You should really consider the total
potential energy EPtot.

> Conformer 1(eq):
> PM3ESCF = -239.3042 (RMS fluctuations 13.0877; Ewald errors
> estimate at NSTEP
> 1000 = 0.3610E-01, at RMS = 0.2741E-02) [I assume these are
> kcal/mol, though
> irrelevant to present comparison; at any event, where to find
> a table of units in use in Amber9?]

Yes, for the record the Ewald error estimate is incorrect in QM/MM
calculations - this is something I need to fix but am not sure how to do it.
For your purposes if it was sufficiently accurate (I.e. you had a big enough
NFFT1,2,3 etc) in a purely classical calculation then it should be good
enough for the QMMM calculation. Sander generally always calculates good
NFFT values for everything but the smallest periodic systems.

The units are KCal/Mol in all AMBER output files. Hence ESCF is in KCal/Mol.

> At this understanding of reading the outputs,I can't reach
> any conclusion as to
> the relative populations.

No in so much that you are trying to do the wrong type of analysis. The
average potential energy may be of use but it really depends on what you
want to observe. A 'traditional' option would be, if you don't want any
dynamic information, would be to minimize the two structures and compare the
potential energies for the minimum's. However, this would be at 0K and would
not necessarily apply at room temperature and of course gives you no dynamic
information. It would also of course contain all the degrees of freedom of
the water etc so you might not get a good minimum.

You could try running cluster analysis (Tutorial B3) for your trajectories -
this would tell you what percentage of your trajectory was spent in what
state, it would also tell you about interconversions between states - I.e.
to get from state X to state Y the system always goes through state Z - you
would likely need longer trajectories than 1000 steps to get good statistics
though and possibly Replica Exchange although I'm not sure on this.

Others may be able to chip in here on other useful options - note the issues
involved here are the same as a classical MD simulation so you can use the
same type of analysis.

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 Wed Aug 01 2007 - 06:07:23 PDT
Custom Search