In simulations, we often use solutes like urea or TMAO and there have been various differing models for these. For instance, in urea we have the Smith models ( for KBFF as well as the one for GROMOS) as well as Duffy model and Amber model. Again for TMAO, we have Kast, Garcia, Netz and Shea models. Often these have very specific bonded and non-bonded parameters as well as combination rules.
There are a couple of doubts as to their incorporation into GROMACS:
1) Errors in minimization often come up ( converged to machine precision but not actually converged) or insufficient parameters error often come up if we follow the models exactly. In particular, the dihedral angles specified for the models are confusing to put in the format of the GROMACS topology file
2) I usually create new atomtypes and define their sigma and epsilon and then specify their bonded parameters in a separate itp file.
3)I find that in several papers, models that were parametrized using geometric combination rule have been used with GROMACS ff like AMBER, that use arithmetic combination rule. How does that stay true to the integrity of that particular solute model then?
Can anyone provide me with a sample itp file or advice about how exactly to solve this problem?
Hi,
http://manual.gromacs.org/documentation/2019/onlinehelp/gmx-solvate.html
Martiniano Bello pointing out solvate doesn't address the question, which is related to topology issues. solvate only builds coordinates.
To address the questions:
1. Not converging below the specified threshold is not necessarily a big deal, but if the forces remain large it indicates a problem with the topology somewhere. A specific example would be helpful. If you have insufficient parameters, that means you're not specifying the parameters correctly given the functional form. Make sure you're using the right function type (e.g. periodic, Fourier, Ryckaert-Bellemans) and use the existing force field files as a guide.
2. Creating new atom types is only necessary if the authors of the models did so themselves, otherwise the typical approach is to make use of existing atom types. Most force fields these days are quite comprehensive, so the introduction of new atom types should be very rare.
3. Without specific references to papers, I can't comment directly on this but it can be the case that the objective of the authors is simply to produce a working model of solvent; it is not necessarily guaranteed to mix with biomolecular force fields. If the combination rules employed are different, it is by definition incompatible and should be viewed as a standalone model.
Thank you Justin,
By new atom types , I meant that for, say carbon of urea, the sigma and epsilon for a particular model would differ from those for standard carbon atom types given in the database. One wouldn't want to give the same atomtypes, as if we're taking protein, then the nonbonded parameters for the carbon atom should belong to that particular ff that we're using.
Apramita Chand urea exists in every force field in GROMACS already. If you're simply trying to use a different version of it, use the existing parameters as a guide for constructing the new ones.
- Badges
- Science topic