Add json files for other UA force fields and lipids

[patrickfuchs]

So far in dic_lipid.py, we have only one UA lipid : Berger POPC. We also have CHARMM36-AA POPC, but it's an AA force field. This lipid is there because it was used to test whether the reconstruction from buildH gave a good match compared to an all-atom force field such as CHARMM36-AA.
In this issue, I suggest to extend dic_lipids.py to :

• other UA force fields, especially CHARMM36-UA. Here's an example of trajectory : https://doi.org/10.5281/zenodo.1293774.
• possibly other common lipids, such as DOPC, DPPC and DMPC.

We shall use autolipmap for building automatically these dictionnaries.

[patrickfuchs]

Samuli Ollila forked buildH and expanded dic_lipids.py for some other UA force fields : NMRLipids@d0a2bc1
So before adding new stuff to dic_lipids.py, we shall restart from Samuli's version.

[HubLot]

We could cherry pick his commit and add it to master through a branch (like this)

[patrickfuchs]

Waoo, very impressive technique 😎 !!!

[HubLot]

I'm on it, should we add also into this repo the others def files from this commit or is it unnecessary?
(I think it's not needed)

[patrickfuchs]

Well, in principle all lipids in dic_lipids.py should have the corresponding def file. So in principle I'd say yes, unless it's just another residue name and we can use the trick Berger_PLA = copy.deepcopy(Berger_POPC).
I'm thinking now, maybe we should find a way to synchronize the def file from the MATCH repo (https://github.com/NMRLipids/MATCH/tree/master/scripts/orderParm_defs) ? This is something we can discuss with Samuli, that'd be nice to have these .def files at a well defined place, rather than randomly trying to edit them from repo to repo.

[patrickfuchs]

I just added Berger POPE. I'm almost finished with Berger cholesterol. I put a check list of the lipids (+ a Zenodo link to a traj example) that would be nice to add (the most important ones are highlighted) :

Berger

• POPC
• POPE
• POPS (https://zenodo.org/record/1129425) <---
• DOPC (https://zenodo.org/record/1402411) <---
• DOPE (https://zenodo.org/record/1293928)
• DLPC (https://zenodo.org/record/16287) <---
• DMPC (https://zenodo.org/record/2605546) <---
• DPPC (https://zenodo.org/record/13934) <---
• cholesterol (https://doi.org/10.5281/zenodo.605429) <---

GROMOS-CKP

• POPC (https://zenodo.org/record/3247435) <---
• POPS (https://zenodo.org/record/1129435)
• POPG (https://zenodo.org/record/3484548)
• POPE (https://zenodo.org/record/1293932)
• DPPE (https://zenodo.org/record/1293957)
• DOPS (https://zenodo.org/record/1129447)
• DOPE (https://zenodo.org/record/1293941)

GROMOS 53A6L

• DPPC (https://zenodo.org/record/14594) <---

CHARMM36-UA

• DPPC (https://zenodo.org/record/17004) <---
• DLPC (https://zenodo.org/record/13821)
• POPS (https://zenodo.org/record/1129458)

CHARMM36 (only for validation purpose since there already H in there 😉)

• POPC

[ohsOllila]

We have now progressed in adding also united atom trajectories in the NMRlipids databank with @larabort and we have come to the point that we would like to add trajectories for which json files are not yet available in here: https://github.com/patrickfuchs/buildH/tree/master/buildh/lipids

Specifically we would be currently interested to add the following lipids simulated with the Berger force field:

1. DOPE
2. DOPS
3. DLPC
4. DMPC
5. DMTAP
6. PAzePC

I believe that it would be straightforward for @larabort to create the json files for these, but how should we proceed in practice in adding those into the BuildH program?

[patrickfuchs]

It is always possible to build your own json file and supply it to buildh with the option -lt. However, if @larabort does this job, I think it could be useful to incorporate them in buildH so that they could be available for everyone. In this case, @larabort can do a PR. A couple of remarks if you'd like to PR though:

• we incorporate new lipids json files which contain all possible Hs to reconstruct ;
• a def file containing all possible C-H pairs to reconstruct should be provided too ;
• the provided files should be checked carefully (one easy test is to reconstruct all H of a single lipid and carefully inspect the lipid with new Hs using Pymol or VMD).
  From my experience, it's easy to make a mistake when writing those files on new molecules without an existing model. When I see the lipids you need, you should be able to use the ones that already exist in buildH as templates. However, for info, we have an automatic procedure which uses autoLipMap. It was very useful for building the files for cholesterol. Of course, we can provide the scripts if needed. In such a case, don't hesitate to contact me.