Bond , angle, dihedral and improper coefficients in large molecules

LAMMPS LAMMPS Mailing List Mirror
1

Dear Lammps experts,

Please let me know, how I can find the bond, angle, dihedral and improper “coefficients” information for all bond, angle, dihedral and improper types in a large polymer molecule.

Actually, I mean is there any software available for calculating these information or not.

by the way, the initial data file has been attached.and I have considered below initial conditions for the polymer molecule.

units real
atom_style full
bond_style harmonic
angle_style harmonic
dihedral_style opls
improper_style cvff

any help is highly appreciated.

Best regards,

Bahman Daneshian, computational materials design, HSU, hamburg

model.lmp (590 KB)

2

Please let me know, how I can find the bond, angle, dihedral and improper “coefficients” information for all bond, angle, dihedral and improper types in a large polymer molecule.

Actually, I mean is there any software available for calculating these information or not.

Yes.
Somebody suggested you should purchase a materials studio license, but I don’t think that is necessary.
(But if you have access to materials studio, by all means use it.)

First of all, for other LAMMPS users who arrive here using google, most people use a molecule builder which creates bonds, angles, dihedrals, and improper interactions and looks up the force-field parameters automatically. There’s a list of molecule builders for LAMMPS that do this at the end of this email, as well as other instructions for generating angles, dihedrals, and impropers, if you have an incomplete DATA file.

However if you simply need to lookup force field parameters for existing angles, dihedrals, and impropers (that are already in the data file) then you have to choose which force field you want to use:

All of the MSI force fields you were considering (CVFF, PCFF, COMPASS) are far from complete, regardless of which molecule builder tool you decide to use. (if applicable… Again, they are listed below.) The force-field files you mentioned (ending in .FRC) intentionally omit many atoms (such as SP2 carbons) because the parameters for these atom types are proprietary and owned by Accelrys. I would imagine it would be difficult to build many common molecules using these force fields as a result (unless you purchase a materials studio license. You might also consider MedeA. I haven’t used either of these commercial products and have no opinion about them.)

The OPLSAA, and AMBER GAFF or GAFF2, and GROMOS force fields do not have this problem, and their force-field parameters are available publicly. (Available in moltemplate format here, here, and here, MSI/.frc format here, in TINKER format here, or in AMBER format here, with documentation here,)

GROMOS 54a7 force field parameters are available here (in gromacs-compatible format)

CHARMM force field parameters (version 27) are available here (again, in gromacs-compatible format). I’m not sure if this version of CHARMM includes 5-body cross terms.

You might also have a look at the force-fields distributed with OpenBabel, including MMFF94, UFF, and GHEMICAL.

------ atom typing -----

For each of the atoms in your molecules, you will have to figure out which of the atom types best describes the atom in your molecule. Knowing that you have a carbon atom is not enough. For example, is it an SP2 aromatic carbon? Or is it a carboxylic-acid carbon? These two carbons will have different names, and this will effect which entry you use in the table of force-field parameters. The files mentioned above usually begin with a list of atom type-names and their descriptions. (Sometimes, as with the “oplsaa.lt” file, the atom-type names are numbers.)

------- symmetry ------

Although you (Bahman) already created a LAMMPS data file containing angle, dihedral, and improper interactions, keep in mind that the number of dihedral and improper interactions which should appear in your molecule could depend on what force field you choose to use. For OPLSAA and AMBER and most other force fields, the symmetry rules for generating dihedral interactions are the same. (Usually they just have one dihedral interaction for every 4 consecutively bonded atoms.) However the number of improper interactions may vary for these force fields. (Perhaps this detail does not matter much since improper interactions are usually only used to enforce planarity.) Furthermore, the MSI class2 force fields (COMPASS, PCFF, and possibly CVFF?) use different symmetry rules for generating both dihedrals and improper interactions. If I’m not mistaken, those force fields will generate more dihedrals and impropers, than OPLSAA, AMBER, or CHARMM. (I listed some symmetry rules I came up with for various force fields here, here and here. I realize these files are difficult to interpret, so I’m not sure this helps at all.)

------- starting from scratch -------

As mentioned earlier, if you are willing to start from scratch, here is a list of molecule builders from the LAMMPS’ web page:
https://lammps.sandia.gov/prepost.html

You can use these tools to build your own molecular systems from scratch. (Not really useful for Bahman, since he already has a LAMMPS data file.)

Some additional tools listed here:
https://sgsaenger.github.io/vipster/
https://sourceforge.net/projects/moleculardynami/
https://github.com/lammps/lammps/tree/master/tools/ch2lmp
https://github.com/lammps/lammps/tree/master/tools/amber2lmp
https://github.com/lammps/lammps/tree/master/tools/msi2lmp

Here’s an old old web page that explains the basic features of the LAMMPS’ DATA file format, in case you want to forgoe using a molecule builder, and create the data file yourself from scratch.

If you are using the OPLSAA force-field (or molecules from the ATB database), then moltemplate (https://moltemplate.org) might be the easiest choice because those force-field parameters are already included with software (or with the ATB file download). (Keep in mind that I am biased, because I wrote moltemplate.) However, for OPLSAA, AMBER, or COMPASS, you will have to manually select the best atom type which describes each of the atoms in your system. AMBER(GAFF, GAFF2) and COMPASS are also included. Moltemplate does not know how to automatically determine atom type from a PDB file.

Tools like EMC might be able to automatically infer atom type from PDB files. (I am not sure about this.) VMD+topotools also might have some ability to interpret PDB files as well. (But I am not an expert using these tools.)

If you are using AMBER, it might be easier try using AmberTools, along with amber2lmp, because AmberTools can attempt to determine atom types from PDB files and calculate partial charges for all the atoms automatically. (With moltemplate, you will have to use an external tool to calculate them and then hand-edit your .LT files to include them.)

If you are using CHARMM, I’m under the impression that most people use ch2lmp tool.

If you are using PCFF or COMPASS, then perhaps you should try EMC (http://montecarlo.sourceforge.net/emc/Welcome.html) or msi2lmp (which might require materials studio). If you are using CVFF, then you are limited to msi2lmp. Note: To my knowledge, neither of these tools (EMC, or msi2lmp) understand the “auto-equivalences” sections of these force fields (PCFF and CVFF). I think this means that some of the angle, dihedral, and improper interactions will be omitted, even though they should be there. I don’t know if a warning message is printed when this happens.
(The COMPASS force field does not use “auto-equivalences”.) Also, as mentioned above, most of the MSI force fields (including PCFF, COMPASS, and CVFF) are incomplete.

bond_style harmonic
angle_style harmonic
dihedral_style opls
improper_style cvff

Theses styles (bond_style, angle_style, dihedral_style, and improper_style) will vary depending on which force-field you want to use. (OPLSAA, AMBER, CHARMM, COMPASS, GROMOS, etc…)

I hope this helps.

Andrew

3

There’s a list of molecule builders for LAMMPS that do this at the end of this email,
as well as other instructions for generating angles, dihedrals, and impropers,
if you have an incomplete DATA file.

I forgot to include that.
(This is not relevant to Bahman, because his DATA file already has these interactions.)

If you have an incomplete LAMMPS DATA file containing atoms and bonds, but lacking angles, dihedrals, or impropers, you can generate those interactions using these tools:

  1. the “guess dihedrals/impropers” functionality in topotools

OR

  1. The nbody_by_type.py utility included with moltemplate (and documented in appendix F.1 of the moltemplate manual, pages 75-77. You don’t need to be a moltemplate user to use the “nbody_by_type.py” program. This script can be run independently from within a bash script which can be invoked from within lammps at regular intervals, although using it this way is somewhat inefficient.)

Either way you will still have to supply the force field parameters for the angles, dihedrals, and impropers which are generated (which is what Bahman was asking about).

Hi

Andrew

My original (long, rambling) reply to Bahman’s question is included below:

4

Dear Andrew,

Thank you very much for your detail response! Actually, although it is very detail, in some cases, it is also a little confusing. Actually, I have noticed that I should check one of the OPLSS, AMBER or Compass forcefulness (since they are open in public). then I should MANUALLY find the atoms, bonds, angles, dihedrals and improper information and ad them to a lammps data file. I have also noticed that OPLSAA is the most similar force file to the CVFF. Am I correct?

Yours sincerely,

Bahman Daneshian

5

Dear Andrew,

Thank you very much for your detail response! Actually, although it is very detail, in some cases, it is also a little confusing.

Sorry. It’s confusing to me too. It is a complicated topic. Perhaps I should have organized my response to your question more carefully.

Actually, I have noticed that I should check one of the OPLSS, AMBER or Compass forcefulness (since they are open in public). then I should MANUALLY find the atoms, bonds, angles, dihedrals and improper information and ad them to a lammps data file.

Yes. You can add them to the “Bond Coeffs”, “Angle Coeffs”, “Dihedral Coeffs”, or “Improper Coeffs” sections of your DATA file.

(COMMENT: Alternatively, you can create a list of “bond_coeff”, “angle_coeff”, “dihedral_coeff”, and “improper_coeff” commands which you would add to your LAMMPS input script. This way you don’t have to modify your DATA file. Some users like Axel and I prefer to use the second approach in order to keep all the force field parameters, including pair_coeffs in a separate file. It does not matter. Either way should work fine.)

…Once you have chosen a force field, then figuring out which atom types to use when looking up these parameters is the next difficult task. (“atom typing”) Once you have figured this out, this should not take you very long. There should not be many types of dihedrals and impropers in your system. For example, the simple alkane-chain example needed only 3 types of angle and dihedral interactions.

I have also noticed that OPLSAA is the most similar force file to the CVFF. Am I correct?

In terms of their behavior, I’m not sure which force field is most similar to CVFF. I’m probably the wrong person to ask.

I can only say that they use different types of equations. It’s true that CVFF and OPLSAA both use standard 12-6 Lennard-Jones non-bonded (“pair”) interactions. However (most versions of) the CVFF force field also include some exotic “class2” (“cross”) terms, such as the “bond-bond”, “bond-angle”, “angle-angle-torsion_1”, and “angle-angle” terms (see the “cvff.frc” file). (Unless I am mistaken, only MSI/Accelrys force-fields have these exotic terms, such as PCFF and COMPASS.) It’s true that PCFF and COMPASS are even weirder because they use 9-6 Lennard Jones interactions (instead of 12-6) and they have even more bonded “class2” terms. In that sense, I suppose they are even more different compared to OPLSAA (or AMBER or CHARMM).

(Complicating matters even further, there are several variants of the CVFF force field which use different equations. See the “cvff.frc” file for details. It has a lot of conflicting information. That’s one of the reasons I didn’t bother trying to figure out how to convert it into moltemplate format. Both PCFF and CVFF are weird and complicated.)

As of 10 years ago, I was under the impression that both OPLSAA and AMBER GAFF/GAFF2 force fields (as well as GROMOS and CHARMM) were extremely popular for simulating small organic molecules (as well as proteins and nucleic acids).

Most people will suggest that you choose the force field you want to use by reading published articles which simulate systems which are similar to your system. (I noticed you have titanium atoms in your system. Is it possible that the molecules are covalently bonded with the Ti atoms? Forgive me. This is outside my area of expertise. I ask because, if so, this will effect the choice of force field you use. If not, then perhaps you can get away with one of the more standard popular force-fields for organic molecules.)

Good luck

Andrew