Re: [AMBER] ff99SB*-ildn again

From: Vlad Cojocaru <>
Date: Thu, 14 Jul 2016 16:45:45 +0200

I think all I wanted to say is that my approach would be to use as many
tests as possible to make sure programs are doing what you want them to
do, no matter which kind of programs. Simply "trusting" is not a good
option in my opinion ... Programs (even those tested carefully) can
have bugs that may manifest only on a specific system because of not
obvious reasons ... In the case of Gromacs versus Amber topologies,
identifying and understanding the unavoidable differences is important,
can be done through testing, and can give confidence that subsequent
simulations are not subject to any input artifacts ...


On 07/14/2016 03:58 PM, Jason Swails wrote:
> On Thu, Jul 14, 2016 at 3:56 AM, Vlad Cojocaru <
>> wrote:
>> Hi Neha,
>> What you are saying argues against trusting conversion tools between
>> Gromacs and Amber (its good you did a thorough test, but many don't do
>> that) ....
> ​I don't think that's what this shows. In fact, I think it would suggest
> precisely the opposite -- i.e., you *should* trust conversion tools that
> convert between topology formats (rather than raw parameter file formats)
> where reasonably possible.
> Ignoring the fact that these are different force fields for a minute, the
> comparison made here is from a GROMACS-generated file converted to Amber
> using ParmEd to a file created directly with Amber's tleap.
> Particularly problematic are how different programs assign impropers --
> they are very often assigned in different orders, and so result in
> non-identical topologies. This is why it's very challenging to get a
> GROMACS-generated file using ff99SB-ildn to agree perfectly with an
> Amber-generated file using ff99SB-ildn.
> By contrast, tools like ParmEd and InterMol take the exact topology defined
> within either the GROMACS or Amber files and converts them faithfully into
> the other format using the exact same atom ordering for every topological
> feature. These tools *are* validated by comparing energies and forces
> calculated by GROMACS and Amber and making sure they agree to the expected
> level of precision (usually ~0.1 kcal or lower, depending on system size
> and precision limits).
> My experience is that I have a lot more confidence in a file converted from
> the file format in which the original FF implementation was proposed into
> another program's format than trusting that the underlying force field
> parameter files were converted correctly *and* that each program's
> respective parametrization tools makes the same choices in all situations.
> In fact, I would use the former to help identify, quantify, and debug
> problems in the latter.
> All the best,
> Jason
> P.S., Of course, if you use different force fields in both programs, as was
> done here, such a comparison is moot, anyway :)

Dr. Vlad Cojocaru
Computational Structural Biology Laboratory
Department of Cell and Developmental Biology
Max Planck Institute for Molecular Biomedicine
Röntgenstrasse 20, 48149 Münster, Germany
Tel: +49-251-70365-324; Fax: +49-251-70365-399
Email: vlad.cojocaru[at]
AMBER mailing list
Received on Thu Jul 14 2016 - 08:00:02 PDT
Custom Search