Is there a possibility that certain atoms (HO) deleted by an update?

GROMACS version: 2022.03
GROMACS modification: No
Here post your question

Dear experts.

I’m a novice user who is trying to generate tpr file of POPG membrane with grompp.
It has been failed because the distributed POPG itp files include HO, which is not included in gromos53a6_lipid.ff, or other gromos43a1~54a7 (while existing in amber99.ff).

HO frequently appears in distributed lipids’ files and those file’s introductions are saying that ‘if you use Peter Tieleman’s lipid parameters (and tutorial), it would work.’
(POPG of the Karttunen Group: SoftSimu - Downloadables: Software, parameters )
(POPG of the Elmore Lab: Untitled Document )

However, I should conclude that it is impossible (although my understanding of GROMACS is not much).
The information of atom (HO) in popg.itp and popg.pdb files are not provided in forcefield.itp. There is no way for grompp to generate tpr file.

So it seems to me that atom HO once had been existed in GROMOS53a6 before, but deleted or replaced at some point. So although it would work before, it would not work now.

In that case, I would start to find and check a previous version of GROMACS.
But I fear that it would be a meaningless effort based on my ignorance.

So I want to ask a question; Is my assume is delusive? Have any atoms been deleted as updates before?

Thanks for your time and attention.


Rather than going into the specifics of the issue, it’s important to say that unless there is a very good reason you shouldn’t use either of the PG lipid parameters you linked to (so those based upon the Berger PC lipids). These parameters don’t behave well (see with very ordered membranes and unusual structures in the glycerol head group.

So I’d definitely recommend using a different PG force field. If you really want to use a united-atom model, instead of all-atom ones like CHARMM36, you can use the GROMOS-CKP PG parameters (available on Lipidbook). These are reasonably accurate and will behave better than the Berger PG models or other united-atom PG models I’m aware of.



Its parameters are understandable and it is working well! Thank you for the good advice!

Dear TomPiggot

Thank you again for your great work!
However, may I ask you a question again?

How could I use a homochiral phospholipid membrane?
If I just delete all DDMG from the pdb file and stabilize it by inflating and re-packing, could I use your membrane as a homochiral membrane?

Or, I just don’t need to delete the D-form phosphatidylglycerol lipid?
Your lipid model contains both of D-phospholipid and L-phospholipid (like many other models…although I don’t know why.).
[Lipidbook Package: GROMOS-CKP Gromacs 1,2-dimyristoyl-sn-glycero-3-phosphoglycerol by Piggot TJ et al.]
As far as I know, those heterochiral lipid membranes are rare in nature and exhibit different properties with homochiral membranes.
If I use the heterochiral membrane, I guess that reviewers would ask why, but I can’t find an explanation yet.

Thanks for your time and attention again.

I’ve probably got an equilibrated homochiral membrane somewhere, I’ll look in my archives. Otherwise you can build your own membrane and equilibrate (e.g. with packmol). As it’s united-atom I imagine you could also just rename the lipids and run an equilibration simulation where the improper that defines the head group chirality would probably flip the state (although you should check that this does happen)

As for membranes in nature, as I’m sure you’re aware, they are typically highly heterogeneous in terms of both head and tail compositions making it very difficult to accurately mimic anyway. So you’re right that you don’t typically get this sort of mix of chirality but you also wouldn’t get a pure POPG membrane either. So no matter what you do, it’s going to be a mimetic, you just need to be able to justify what you’re doing.

Thank you for advice!

So I can just rename D-form lipids of .gro or .pdb file to L-form lipids because chirality information is encoded in itp. It is easy!

But, although it is really embarrassing me, may I ask you one more question?

Your files allow me to pass grompp (the step that obstacle to me for 1 month). Although some warnings appeared, It was much better than previous trials.

But after I checked the box with VMD, I found a weird result.

After merging peptides and DMPG and filling the water in the box, there was no DMPG. Only water and peptide were exhibited.

While the grompp command (gmx grompp -f minim.mdp -c DMPG.pdb -p -o DMPG.tpr) working, there were three warnings. I guess that one of them would have cause this phenomenon.

I guess that most suspect warning is:

WARNING 1 [file DMPGL.itp, line 271]:
  Too few parameters on line (source file
  /gromacs-2022.3/src/gromacs/gmxpreprocess/toppush.cpp, line 1900)

Because one of the other warning

WARNING 2 [file, line 27]:
 The GROMOS force fields have been parametrized with a physically
 incorrect multiple-time-stepping scheme for a twin-range cut-off. (the rest omitted)

is usually not cause a serious problem.

And the other one

atom name 16598 in and DMPG.pdb does not match (NA - NA1)
atom name 16598 in and DMPG.pdb does not match (NA - NA1)
atom name 16617 in and DMPG.pdb does not match (NA - NA1)

(more than 20 non-matching atom names)

WARNING 3 [file, line 27]:
  128 non-matching atom names
  atom names from will be used
  atom names from DMPG.pdb will be ignored

disappeared after I changed ‘NA1’ in .pdb file to ‘NA’ (Actually, it is also somewhat weird. Sodium in your .gro files is named as ‘NA’. But it changed to ‘NA1’ in .pdb after I use the command (gmx editconf -f DMPG.gro -o DMPG.pdb)…well it is not important for now.)

According to my short search, ‘Too few parameters on line’ is the warning that appeared when I used the force field unmatching with .itp file. But I included gromos53a6.ff in file.
Could I ask you again for your advice on what’s wrong?

So it looks like there are several different things going on based upon your output.

But before getting into these specific issues, you need to be both careful and sure of what you’re wanting to do.

For one, if you do change the D lipids to L then you must check that this changes the isomeric state of these lipids. You are right that the itp (as its united-atom) does define the isomer but you need to check that the structures change upon applying the new topology. Given it’s just the head group I imagine they will but you need to check this.

Beyond that, I’d suggest that for your system you should probably use all atom force field like CHARMM. That’s unless there is a very good reason not to. If there isn’t then using the CHARMM-GUI will allow you to setup your system easily for GROMACS.

So if there is a good reason to not use CHARMM force field please say, otherwise that’s the easiest way forward for you to overcome these issues.



Thanks for the advice!
Unfortunately, it seems that the access to charmm-gui from my country is currently blocked because of violating several cases of the user agreement by other users.
But your advice about using all-atom is really helpful. I didn’t know about charmm-gui and it seems like good tool.

So I would study the method to use CHARMM ff in gromacs for now. It seems that I need to read more tutorials and manuals and advice to use it because CHARMM ff didn’t work with the same steps as GROMOS did.

or I would be able to access to charmm-gui from the USA (I would go to the USA within month).

And there was no particular reason to not use CHARMM. I didn’t used CHARMM force field, because I understand that the abstract in the lipidbook is recommending GROMOS 53A6 and GROMOS 54A7 force fields.