On Mon, Oct 7, 2013 at 9:58 PM, Robert Molt <rwmolt07.gmail.com> wrote:
> Sirs and Madames:
>
> I am having a confusion on Amber Tutorial B1. In the generation of the
> .prmtop file associated with vacuum DNA, I observe:
>
> %VERSION VERSION_STAMP = V0001.000 DATE = 10/07/13 21:15:38
> %FLAG TITLE
> %FORMAT(20a4)
> default_name
> %FLAG POINTERS
> %FORMAT(10I8)
> 638 14 234 452 554 694 1162 1238 0
> 0
> 3544 20 452 694 1238 32 59 39 24
> 0
> 0 0 0 0 0 0 0 0 33
> 0
> 0
>
> I have bolded the relevant number. The website lists the correct set of
> values as
>
> 638 14 234 452 554 694 1056 1238 0
> 0
> 3544 20 452 694 1238 32 59 39 24
> 0
> 0 0 0 0 0 0 0 0 33
> 0
> 0
> The correct value is 1056, my value is 1162. My PDB file matches the
> website-posted pdb file. I am using ff99SB, which I believe is a unique
> specifier. Admittedly I am doing all of this through tleap, rather than
> following xleap in the tutorial, but this should not matter in principle.
>
> Can anyone advise what mistake I am making that results in this difference?
>
Your mistake is putting too much stock in the provided files ;). I used
ff99SB to generate the topology file and got the same pointers you did in
the head of the topology file. The difference is the number of torsion
terms including hydrogen. Torsion terms have a tendency to change from one
force field to another since many torsions are modeled with multiple
torsion terms (think Fourier series describing the periodic potential), as
well as additional 'improper' torsions that are meant to keep 4 atoms
coplanar (e.g., in an aromatic ring).
I suspect that the force field used to generate the topology file provided
in the tutorial is slightly different than the one that is produced using
"ff99SB" today. Consequently, I would recommend using up-to-date force
fields when trying the tutorials (ff12SB is what I would recommend), and
don't worry about the topology files not matching exactly. These tutorials
are a challenge to keep up-to-date with the latest-and-greatest force
fields, so I would not be surprised if a number of topology files turned
out to be different like you have observed here. (It's not enough to
simply update the generated prmtop -- you then have to update all of the
output files generated in the tutorial.)
Reproducibility out to an arbitrary level of precision is fairly important
to QM calculations as I understand it (HF/6-31G* should give the _same_
answer for the same input regardless of what program you use).
Reproducibility in this field is far less concrete. While sometimes
important (for validating a force field implementation, for instance), two
simulations are sometimes not directly reproducible even with the same
program and the same inputs on the same machine due to round-off errors and
the chaotic nature of classical interactions between 3+ bodies. While
steps can be taken to ensure that 2 simulations are 'identical', this is
not practically useful for running simulations, [1] as only ensemble
properties are of any interest in MD simulations.
HTH,
Jason
[1] The only use I can think of for perfectly reproducible simulations is
to aid in debugging.
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Oct 08 2013 - 05:00:04 PDT
