Dear Amber users,
Please help me. I cannot perform the local optimization of the energy of
a molecular system, e.g. a protein-ligand complex, calculated in the
frame of the Amber force field in vacuum with fixed protein atoms and
with variations of ligand atoms.
I try to compare different force fields by their ability to perform
docking. Such comparison can be made by the quasi-docking procedure as
follows (see
https://doi.org/10.1155/2017/7167691. We performed docking
in the frame of the MMFF94 force field with implicit solvent by massive
local optimization (in average about 10^5 local optimization per
complex) of the protein-ligand energy with a rigid protein and with
variations of coordinates of ligand atoms. We obtained 8192 low energy
minima for each of 16 test protein-ligand complexes. Docking of native
ligands into their own proteins was performed for complexes which were
taken from Protein Data Bank. The low energy minima search was almost
exhaustive, i.e. practically all low energy local minima have been found
for each protein-ligand complex. The landscape of the MMFF94 energy
surface is reflected by this set of low energy minima. It is possible to
compare the ligand pose corresponding to the global energy minimum with
the crystallized native ligand pose in the complex. We want to compare
MMFF94 and Amber force fields in docking by recalculating the energy of
each minimum (obtained in the MMFF94 force field) in the Amber force
field. We have to perform local optimization of the energy of the
protein-ligand complex in the frame of the Amber force field. The ligand
initial pose in each of these optimizations will be each of 8192 energy
minima obtained in the MMFF94 force field. Certainly the energy must be
calculated with an implicit solvent model. The main problem now is the
local optimization of the energy of the protein-ligand complex in vacuum
following by the minimum energy recalculation with a solvent model
because the optimization with solvent model is too slow – we have to
perform the optimization of each of 8192 minima for 16 test complexes.
Thus, we met problems at the stage of local energy optimization with
Amber tools (Sander) in the frame of the Amber force field in vacuum.
1.Optimization with gradient methods (ntmin = 0-2). The protein is rigid
and coordinates of all ligand atoms are varied. The optimization does
not stop at the minimum configuration! How is it possible to perform
careful optimization with this method?
2.It is possible to use the XMIN (ntmin = 3) method or the LBFGS method
(xmin_method = 'LBFGS') but the optimization is finished after several
first steps with ERROR in load_lbfgs(): YS=0. How is it possible to
avoid this error?
3.Other methods of the local energy optimization in XMIN do not perform
properly: either the local energy minimum pose is slipped or the
optimization jumps between two points with different energies for too
long time.
4.We tried to perform the energy optimization with the Amber force field
with other packages: NAMD and Charmm. In these cases the energy
optimization finished well. However, when we continue the energy
optimization with the Sander module we saw that the minima found by NAMD
and Charmm are not minima in fact!
So, can anybody explain me how to perform good quality optimization of
the energy of the protein-ligand complex (the protein is fixed and
coordinates of all ligand atoms are varied) in the frame of the Amber
force field in vacuum by tools of the Amber package?
Thank you in advance.
Danil Kutov
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Apr 01 2017 - 01:00:02 PDT
