Lincs and electrostatics

User discussions
1

Hi all,

I have a couple of questions;

  1. What is the purpose /advantage of having Lincs constraint on H bonds only, and why not other bonds?

This is part of the NVE ensemble of the gromacs tutorial
constraint_algorithm = lincs ; holonomic constraints
constraints = h-bonds; bonds involving H are constrained

  1. What is the difference between the PME and cut-off for long-range electrostatics? Why is PME usually preferred? How much do they differ in terms of the cost of the simulation?

The cutoff is used in the gromacs tutorial for adding ions, but for the next steps, the coulombtye is set to PME.

2

This depends on the force field. Most are parametrized to constrain only bonds to H atoms. You need to follow the convention for the force field you’ve chosen.

A plain cutoff is wildly inaccurate and should never be used. PME is the de facto standard for accurate condensed-phase simulations.

This is done to avoid a warning from grompp about using PME with a charged system, and the .tpr for adding ions is never used for any physical simulation so it does not have to even have sensible settings.

3

Thanks, Dr. Lemkul,

The forcefield I’m using is Charmm36 for protein simulation. I want to know what the reason is for restraining H-bonds only. Or why are we not restraining other bonds like C-C or C-N?

Thanks

4

Please note first that “constraints” and “restraints” are very different concepts. We’re talking about constraints here.

Bonds between heavy atoms are not constrained in CHARMM because that is the force field convention. They are not fixed at ideal lengths; they are allowed to vibrate according to a harmonic potential. You have to use this convention properly, otherwise you imbalance other bonded terms (angles and dihedrals) that have had their energy surfaces computed under the assumption that the constituent bonds are vibrating and not held fixed. Bonds to H are constrained to enable a 2-fs time step.

5

In addition to Justin’s answer: constraining all bonds is incompatible with GPU-resident mode, so constraints = h-bonds can be much faster when run on a modern GPU.

But performance is secondary after correctness; don’t use h-bonds where all-bonds is needed :)