TIP4/5P virtual sites - ptype and charge in charmm36-jul2022.ff

GROMACS version: 2022.3
GROMACS modification: No

Hello,

potentially a silly oversight on my part, apologies in advance, but better safe than sorry, they say…

In the latest CHARMM36 force field for GROMACS, charmm36-jul2022.ff, the virtual sites relative to the TIP4P and TIP5P water models (MWT4 and MWT5, respectively) have pytpe = A (i.e., they are labelled as “atoms”). Shouldn’t they have ptype = D (virtual sites - incidentally, in the GROMACS 2022.3 reference manual, particle type section, Table 10, “virtual sides” should probably read “virtual sites”)?

In addition, neither MWT4 or MWT5 have any charge associated with them - they should, however.

In the 2021 versions of the CHARMM36, MWT4 and MWT5 were labeled as virtual sites and they did have a non-zero charge.

Am I missing something? Again, apologies if this is me failing to appreciate some detail.

Best wishes,

Gabriele

We will fix these for a future release, but in practice, I think the errors would only affect water topologies directly written by pdb2gmx, which would be extremely uncommon (since pdb2gmx cannot generate virtual sites anyway). The individual TIP4P and TIP5P .itp files we provide are correct in terms of the charges (the values in ffnonbonded.itp are never used, so whatever you see there is irrelevant) and virtual site construction.

Note that TIP4P and TIP5P are not formally part of the CHARMM force field, but we provide these parameters in case people want to test things. Hence they do not undergo any kind of rigorous testing for implementation. They’re a convenience only.

@awacha can we look at this to get it fixed for a future release of the force field port?

Hello,

thank you for getting back to me so quickly!

I appreciate that CHARMM was originally parametrised with TIP3P “in mind”, but I am rather confident in saying that there are quite a few GROMACS users out there that are using TIP5P and even more so TIP4Pas well - particularly in light of the fact that these water models seem to be behaving quite accurately in conjunction with CHARMM (even though the latter was not parametrised with the former in mind).

I also appreciate that, in most cases, users would not go through the pdb2gmx route to work with TIP4P.

It’s just that some of the TIP4P parameters were wrong in the charmm36-jul2021.ff already - it’s only a few lines, and it’s not a great effort for a user to check every release manually (as you can tell, I always do that in any case…) but it’d be great to fix them nonetheless!

Thank you for your help, massively appreciated indeed.

Best wishes,

Gabriele

As a bit of background, the reason why stuff like this is sometimes incorrect is because it has to be manually added. We completely overhauled the conversion utility last year (wrote a new one) that is vastly superior to what we used to use, but TIP4P and TIP5P, being non-standard for the force field, do not come from an automated conversion. They are boilerplate lines that need to be added. We will get this fixed and thank you for pointing it out.

Hi Gabriele and Justin,

thanks for your report and sorry for the late reply. I’ve fixed the conversion utility. Now partial charges are included in ffnonbonded.itp where [atomtypes] are defined, and the particle type of dummy atoms has been set to “D”. The TIP4P LJ parameters should also have been fixed. I hope that we can push out the updated force field tarball in the near future.

Cheers,

Andras

Hello Andras,

thank you!

Best wishes,

Gabriele