Re: [AMBER] questions about paramfit and frcmod file

From: David Cerutti <>
Date: Mon, 23 Mar 2020 02:36:44 -0400

If you remove certain terms form the frcmod that were necessary to complete
the topology, then tleap will tell you that it couldn't find the
information it needed (you took it away). There is probably a way in
paramfit to mask out certain dihedrals, but I can only tell you about mdgx,
which is the program that Amber will be putting forward for optimizing
dihedral angles starting in Amber20.

The way to do this in mdgx is to make a trajectory, with all the
configurations of your molecule, a prmtop for that trajectory, and a file
listing the QM energies of each configuration, one per line. You need only
make the prmtop once, so long as you are not changing atom types and you
are merely adjusting the stiffness constants of the dihedrals.

Then, you will make a file something like this:

  -parm <your Amber home directory>/dat/leap/parm/gaff.dat
  -d carbofluid.dat
  -o fit.out

  System Glycerol/coords.cdf Glycerol/energies.dat
  System MPD/coords.cdf MPD/energies.dat
  ParmOutput frcmod
  eunits hartree,
  accrep report.m
  verbose 1,

  % Angle fitting input
  fita c3 oh ho
  fita c3 c3 oh
  fita c3 c3 c3
  FitAnglEq 1,
  arst 0.0002,
  arstcpl 1.0,

  % Torsion fitting input
  fith c3 c3 c3 c3
  fith hc c3 c3 oh
  fith hc c3 c3 c3
  fith oh c3 c3 oh
  fith h1 c3 c3 oh
  fith X c3 oh X
  fith c3 c3 oh ho
  hrst 0.0002,

In the above, which is taken from the mdgx tutorial here,, I have give it two
trajectories for two different systems that contain instances of the same
parameters. I then enumerated the parameters with the "fith" keyword
(alias FitH, meaning Fit DIHEDRAL). There is also fita, which is Fit
ANGLE. The four atom types for the relevant dihedral follow. The hrst
setting of 0.0002 is a good value to start with--this is the normalization
towards zero, a harmonic restraint imposed on each parameter to penalize it
having any value at all. The setting of 0.0002 will have the same effect
on your results no matter how big your data set--it scales with the number
of data points and the number of places each parameter appears in them.

What you'll do is make your trajectories, energy lists, and prmtops, then
run mdgx on input like this to get a new frcmod file (the file name is the
-d keyword up in the &files namelist). Then, it's like sander, mdgx -O -I ( is your input file), and read through the fit.out to see
what mdgx thinks of your data set. If you want to optimize thirty
dihedrals that's a lot, but mdgx will let you know if your data set is up
to the job and help you expand it if that's what you really want to do.


On Sun, Mar 22, 2020 at 6:11 PM ttz <> wrote:

> Dear all,
> I am currently trying to develop GAFF for cefotaxime using “paramfits”.
> There are more than 30 dihedral angles need to be optimized, so I
> decided
> optimized the dihedral angle parameters of one subgroup first(10
> dihedral
> angles). After finishing this part, I want to keep those optimized
> parameters and then optimize another subgroup. So I change the
> parameters in
> the frcmod file to generate prmtop file, but there are errors saying
> that
> “ Error: ** No torsion terms for cc-cd-cf-n2, Error: ** No torsion
> terms
> for cc-ss-cc-cd”. However, there parameters do exit in the frcmod file.
> Could anyone help with this? I have attached the original and modified
> frcmod file, as well as the mol2 file.
> Thanks in advance!
> Best regards,
> Tingting
> Sent from [1]Mail for Windows 10
> References
> 1.
> _______________________________________________
> AMBER mailing list
AMBER mailing list
Received on Mon Mar 23 2020 - 00:00:02 PDT
Custom Search