Hi all,
I use Gromacs for protein-water simulations (protein with multiple
chains). I use CHARMM27 with cmap forcefield and tip3p water. Parts of
my chains are position restrained.
I am considering shifting to lammps and want to know if this is
advisable. I would really appreciate any advice on the same.
Especially on the feasibility and convenience.
Regards
Pooja
Hi all,
I use Gromacs for protein-water simulations (protein with multiple
chains). I use CHARMM27 with cmap forcefield and tip3p water. Parts of
my chains are position restrained.I am considering shifting to lammps and want to know if this is
please let us know what it is in LAMMPS that makes
you wanting to move away from Gromacs?
advisable. I would really appreciate any advice on the same.
Especially on the feasibility and convenience.
LAMMPS currently doesn't support CMAP corrections,
so before switching to LAMMPS, you would have to
implement (and test!) those.
axel.
I am using GRomacs with the Plumed plugin for metadynamics which
requires double-precision GROMACS. The double-precision version is
very slow and I do not think I can continue to run some of my more
expensive simulations using it. Hence the switch.
Regards
Pooja
I am using GRomacs with the Plumed plugin for metadynamics which
requires double-precision GROMACS. The double-precision version is
very slow and I do not think I can continue to run some of my more
are you sure that this is due to compiling in
double precision? i've never seen that. double
precision gromacs is somewhat but not massively
slower than the single precision version. i would
suspect that it is rather due to compiling w/o the
SSE kernels (which are indeed slow, particularly
the ones written in C), which may be a side effect
of using plumed.
expensive simulations using it. Hence the switch.
but how should LAMMPS help there? for the force
field kernels that Gromacs supports, it is even slower.
and there are no special optimizations for rigid water
molecules either.
axel.
To be blunt, if convenience is what you want, and if you are just
interested to simulate a "protein-in-a-box-of-water", then LAMMPS is
not the easiest program to use. Other programs are stronger in that
department. To use LAMMPS, my impression is you currently have to use
other tools to set up the simulation (like amber's leap/antichamber,
or gromac's pdb2gmx, genion, grompp, or materials studio), and then
convert those files into LAMMPS format later. From what I hear, this
conversion is the main problem. Converting PSF topology files into
LAMMPS is not difficult (thanks to topotools). However I keep reading
complaints on this mailing list about other automated scripts
(amber2lmp, msi2lmp) which apparently are difficult to use and/or may
be out of date.
On the other hand, Lammps is wonderful for coarse-grained biopolymer
models, because it has a huge diversity of force-fields you can use,
and because you can define new force-fields. For this, you must
construct input file in LAMMPS data/in format.
marketing buzz incoming...
In my free time, I have been working on a LAMMPS preprocessor
("moltemplate/ttree") for defining molecules and using canned force
fields (it supports some rudimentary coordinate generation). It was
designed to construct coarse-grained simulation input, but when ready,
I imagine it may also be possible to port all-atom popular
force-fields (eg. GAFF, amber99) into this format. (More importantly,
it should be possible for force-field files to store parameters for
LAMMPS-specific or user-defined potentials.) I don't feel comfortable
to post it in it's current disorganized state to the mailing list for
all to see, but I'm happy to mail it to individuals who are brave and
have a strong stomach.
If you are just interested in LAMMPS because you think it is faster
then gromacs, then you are probably not the target audience (and I'm
not sure lammps is faster). I hope this helps. (Please excuse the
plug.)
Cheers!
Andrew