LINCS constraints all bonds to h-bonds only

Hi everyone,

In the mdp files on the github pages of pmx benchmarks, LINCS constraints are always set to ‘all-bonds’. However, this causes a significant hit in performance on GPUs as the -update gpu flag to mdrun cannot be used. I think that in the Amber99SB-ILDN paper, constraints are also set to h-bonds only. Is there a specific reason as to why LINCS constraints are always set to ‘all-bonds’ in pmx? Is it “safe” in terms of accuracy to set the constraints to h-bonds for simulations done with the Amber force field only?

Kind regards,


setting constraints to h-bonds should be safe in this case as long as you limit the time step to 2 fs or less

1 Like

Great, thanks!

Also, make sure to use ‘develop’ branch of pmx and do not allow for hydrogen-to-heavy_atoms morphs.

I can understand why hydrogen-to-heavy_atoms morphs are not allowed, but why is it necessary to use the develop branch?

It is the branch with the latest updates

Are there any specific reasons as to why the master branch will not work? Also, is it now recommended to use the develop branch, as the github page still mentions it is not a stable/reliable release?

Amino acid mutation libraries in the ‘master’ branch allow hydrogen to heavy atom transformations, therefore LINCS has to be set to all-bonds.

1 Like