There are several tools for building geometries and databanks available that provide structure coordinates for molecules.
Some possible tools to help you build your molecule are VMD and Accelrys which would allow you to visualize your structure and I think they have some building tools.
An example of a databank is the RCSB Protein Data Bank (
Dear All,
Thanks Michael, for your comments. Yes, material studio (Accelrys) can be used to build polymers from drawing a repeat unit. I have used it together with Lammps for MD simulation on normal polymers.
But I am afraid proteins can’t be constructed from a repeat unit like normal polymers (maybe I am wrong).
Protein Data Bank seems one of the good options. But I am not very clear about one thing: the proteins provided by this data bank are very specific in structures, for example, for actin filaments, there may only provide data for 2 molecules with certain segment in each molecule. It seems we have little flexibility to study other structures for actin. Please correct me if I understand it wrong.
Thank you for your time and help,
Lili
Dear All,
Thanks Michael, for your comments. Yes, material studio (Accelrys) can be
used to build polymers from drawing a repeat unit. I have used it together
with Lammps for MD simulation on normal polymers.
But I am afraid proteins can't be constructed from a repeat unit like normal
polymers (maybe I am wrong).
it is similar, but a bit trickier, due to improper dihedrals for the
backbone. it may be possible that msi2lmp can handle it, but i never
tested it. it would be simple to try this on an alanine dimer and
manually verify the result.
apart from that, there are several tools that can generate coordinates
for aminoacid chains from a given sequence and angles. those can then
be subjected to the regular topology building tools. the ones that i
am familiar with, are the ones shipped with VMD, but there are others,
too, that are bundled with other "bio" MD packages. LAMMPS is a bit
lacking in this area, but improvements are under way.
http://www.ks.uiuc.edu/Research/vmd/plugins/molefacture/
http://www.ks.uiuc.edu/Research/vmd/plugins/autopsf/
http://www.ks.uiuc.edu/Research/vmd/plugins/psfgen/
once you have the topology, you need to convert it into a suitable
LAMMPS data file.
https://sites.google.com/site/akohlmey/software/topotools
Protein Data Bank seems one of the good options. But I am not very clear
about one thing: the proteins provided by this data bank are very specific
in structures, for example, for actin filaments, there may only provide data
for 2 molecules with certain segment in each molecule. It seems we have
little flexibility to study other structures for actin. Please correct me if
I understand it wrong.
you are likely not the first person studying this kind of system, so
the best way to decide on a suitable strategy would be to survey the
literature and look at what other people did and then look for tools
that can do something similar enough, or - if needed - write something
your self. there are several "modeling toolkits" around that allow
very flexible "molecular programming" without having to learn too much
programming, e.g. some python or similar scripting may suffice.
axel.
Dear All,
Thanks Axel for your informative comments. There is another point not very clear to me regarding assigning the force filed to the structure I have by using VMD and NAMD. I know this question is not directly related to Lammps, but I hope I can get some insight here 
If I use VMD and NAMD to run MD simulation, the force field parameter file is required. My question is how can I get such parameters? Should I look up the force parameters by myself from literature or there is a plug-in to do this for my structure (like in Material Studio, the force parameters are assigned once I select the type of it)?
Thanks for your help,
Lili
Dear All,
Thanks Axel for your informative comments. There is another point not very
clear to me regarding assigning the force filed to the structure I have by
using VMD and NAMD. I know this question is not directly related to Lammps,
but I hope I can get some insight hereIf I use VMD and NAMD to run MD simulation, the force field parameter file
is required. My question is how can I get such parameters? Should I look up
the force parameters by myself from literature or there is a plug-in to do
this for my structure (like in Material Studio, the force parameters are
assigned once I select the type of it)?
to build a topology, you need to know which "force field" you want to
use. people use the term force field for different things, but in your
context a force field is a set of rules how to assign atom types,
bonds, angles, dihedrals and their and how to match those types to
parameters. those parameters are usually stored and looked up from
lists or tables. in addition, especially for the case of the
simulation of biopolymers, there are often lists of "templates" that
can be used to automatically assign types, topology information,
parameters and even reconstitute missing (hydrogen) atoms. many of
these are very similar for different force fields, some are different,
and there are always exceptions to the rules.
as far as LAMMPS is concerned, you have to be aware that different MD
codes often put different focus on different applications and thus as
a consequence they work differently well or are easier or harder to
use for different problems. when developing software, people have to
make choices and different choices lead to different programs, that do
- in principle - the very same thing. for biopolymers, there is thus a
whole ecosystem of codes/tools and force fields. typically, the older
codes were developed alongside and to implement a specific force
field. more modern codes support multiple force fields, either in
their original table format or through converter tools. LAMMPS is very
much on the extreme side of being flexible and supporting many
functional forms, but it doesn't have lots of fancy tools to make this
process as easy as it could be. this needs people that have an
interest in it sit down and program them, most end up rather using a
package that has the tools already.
axel.
Most people who want to simulate these kinds of systems do not use
LAMMPS. (They use AMBER, GROMACS, NAMD, or perhaps OpenMM, or other
newer alternatives.)
I've been trying to resist advertising the program I wrote
("moltemplate"), because I'm not sure it will do what you want. I
wrote "moltemplate" to build simple coarse-grained protein models in
LAMMPS. I did not plan people to use it to build all-atom systems.
However I have recently made the AMBER GAFF force-field database
available to moltemplate users. (This is an experimental feature.
This means moltemplate users can build simple organic compounds
without having to know the force-field parameters for their molecules.
They just have to give each atom the correct type name.) I think I
can also convert more recent protein-specific AMBER force-fields into
moltemplate format (to make them available) as well, such as ff99SB,
ff99SP, ff03ua, ff03SB, (Unfortunately force-fields requiring
polarizability or GBSA solvent can not be made available unless
somebody is willing to implement those new features in LAMMPS.)
Unfortunately, to construct an all-atom protein (using moltemplate),
somebody would still have to manually define all 20 amino acids in
moltemplate format (including hydrogen atoms), using AMBER-complaint
atom names. You would also have to assign (partial) charges to each
atom. Features like reading PDB files (parsing different atom-name
conventions), adding missing hydrogen atoms, and solvating proteins
are also still not implemented.
As Axel said, different simulation packages like AMBER have their own
tools for building molecules and asigning force-fields which are
usually specific to that particular simulation package. For amber
(the software I am most familiar with), these tools are free
(http://ambermd.org/AmberTools-get.html) and have tutorials
(http://www.rosswalker.co.uk/tutorials/amber_workshop/Tutorial_three/section3.htm).
These tools will create a PRMTOP file, which must be converted into
LAMMPS format using amber2lmp (which apparently has some bugs which
affect the improper-angles.)
Perhaps it's easier to use some other program besides LAMMPS which has
better tools for building all-atom solvated proteins? I'm sure there
are tools for GROMACS and NAMD (in the later case, using VMD) as well,
but I barely know anything about them.
I don't think my post on this topic tonight was very helpful, but if
you want, I can send you more information about the AMBER force field
features available in moltemplate.
Cheers
Andrew
Hi again. I'm a little bit embarassed because bringing up moltemplate
was inappropriate for all-atom proteins.
I'm not sure what the best free tool is for simulating all-atom
proteins, but instead of using LAMMPS, I would consider using a
simulation engine deticated to that task (only because the tools which
come with those programs are more fully developed). Somebody in my
former lab told me about OpenMM. Like LAMMPS, OpenMM is
free/open-source, and modern, and somewhat modular, and written in
C++. It can definitely read PDB files and it automatically assigns
force-field parameters for many different many force-fields. (And I
think it has a beautiful python scripting interface, although you may
not care about that.) It has a team of some hi profile research
groups supporting it.
There are some programs which can be used to build the input scripts
which OpenMM reads here:
http://builder.openmm.org/
The VMD tools which Axel recommended could be very helpful to for
preparing some of the input files (eg PDB files) which OpenMM reads.
(Also see avogadro.)
I hope this post is more helpful
Cheers
Andrew
Hi All,
Thanks Andrew and Axel for your comments. They are really helpful to me, since I am just starting a new project on MD simulation and multi-scale modelling on neuron cells.
I am figuring out the proper strategy to do MD simulation for proteins, for example, get the protein structure, force field selection, and simulation software,…
VMD (maybe + NAMD) seems a good choice to prepare the protein structure with/ without water. The files generated by this can be used to run simulations by other softwares.
Now, I am thinking about how to assign the force field parameters in an easy way (great to know OpenMM can do this automatically). I will check this out soon.
Thanks again for sharing and help,
Lili
Hi All,
Thanks Andrew and Axel for your comments. They are really helpful to me,
since I am just starting a new project on MD simulation and multi-scale
modelling on neuron cells.I am figuring out the proper strategy to do MD simulation for proteins, for
example, get the protein structure, force field selection, and simulation
software,...VMD (maybe + NAMD) seems a good choice to prepare the protein structure
with/ without water. The files generated by this can be used to run
simulations by other softwares.Now, I am thinking about how to assign the force field parameters in an easy
way (great to know OpenMM can do this automatically). I will check this out
soon.
anything automatic comes with its drawbacks. either it will be low
quality or easily faulty. most force field parameter and atom type
assignments are based on heuristics and matching (parts of) a file to
a template database. but for that to work, you need to closely follow
the expected conventions.
in short, there ain't no escape from the blues. if you want it to be
done well, you have to know (and learn) the gory details. any attempt
at cutting corners will cost you double and triple later.
axel.