drymartini and martini file issues

[piskuliche]

Hello!

I've been playing around with the dry martini model incldued with moltemplate - I had an error that certain atom masses and atom types were undefined. It seems like this issue cropped up especially surrounding the small bead naming schemes. In particular, throughout the lt file there were references to SP4_bSP4_aSP4_dSP4_iSP4, but the replace command read replace{ @atom:SP4 @atom:SP4_SP4_aP4_dP4_iP4} missing S's before the P's in the longer name causing moltemplate to think there weren't masses defined for these atoms.

There was also one bond coeff not defined - P1-C1, I wasn't sure whether that should be defined or not - I added a coeff for it and marked the location (line 828). The coeff I chose was 1.49375 4.8 (which I think is correct).

I've gone through and corrected (I think) all of these issues - I don't make any promises that anything else in the file is correct. (I marked things I changed with comments). Moltemplate runs with dry martini now at least with these changes.

I've attached the new lt file to this issue report (I had to swap the extension to get github to accept it). Let me know if you have any questions, I hope this helps!

Best,
Zeke

drymartini.txt

[piskuliche]

Hello,

Sorry for the second message but I found another issue I think. It seems like the pair coefficient sigmas in the moltemplate force field file for dry martini is still in units of nm rather than Angstroms (like is used for units real in lammps and defined at the bottom of the force field). If I manage to make a corrected version with this change, I'll post it here.

[jewettaij]

Hi Zeke!
The "drymartini.lt" file was created by David Stelter back when he was a postdoc in Tom Keyes' lab (also at BU). I went ahead and included his version of "drymartini.lt" with moltemplate, but I haven't used MARTINI or DRYMARTINI yet. So I rely on people like you to find the problems with these files, and I'm very grateful.
I am a little bit embarassed by the sorry state of the MARTINI (and DRYMARTINI) examples.
In the original ITP files posted on the cgmartini.nl website, each molecule type (for example, each type of lipid) comes with a completely independent list of bonds and angles. In the files from the cgmartini.nl web site, the @Angle type is not one-to-one (not uniquely determined) the @atom types. Most of the time, the type of angle used (@Angle) can be determined uniquely by the types of the 3 atoms (the three @Atoms).

Unfortunately, instead of converting those files, instead we copied our MARTINI and DRYMARTINI force field parameters from the "PRM" files that are distributed with the EMC software. The EMC software requires that angles are looked up according to atom types. So the @Angle must be one-to-one with the @atom types. Whenever EMC discovers a new molecule or lipid type where the same atom types being associated with a different @Angle type, it creates a new @atom type, which is only used in that kind of molecule. This only happens in a minority of the lipid types, but it is an issue you need to be aware of.

Lately, I've been thinking about starting over, and writing a converter that converts the simple ITP files that are on the cgmartini web site (for example, download the ZIP file for DRYMARTINI here. (I feel bad doing that because I know David Stelter put a lot of work into his EMC converter tool.) But the ITP file format used at cgmartini.nl is simple to understand and elegant. (At least, the ITP files for MARTINI and DRYMARTINI are simple. I don't know if GROMOS files are more complicated.) In the long run, I think writing a simple ITP-to-Moltemplate converter would actually be easier than trying to figure how to fix the problems with the "martini.lt" and "drymartini.lt" files derived from EMC (and figure out what EMC did to the atom type names). And we could get the force field parameters straight from the source (cgmartini.nl), and update them to the latest version of MARTINI. (The new Martini 3.0 parameters just came out a month or so ago, but "martini.lt" is still using version 2.1.) Perhaps I should remove all of the MARTINI and DRYMARTINI files and examples from the moltemplate distribution until this is done.

If you plan to use DRYMARTINI or MARTINI in LAMMPS have any time to devote to this, I would probably spend it doing writing that conversion tool. It will probably be 4 or 5 months before I can devote any time to this.

If that's too much work for you, but you find a way to edit the existing "drymartini.lt" file so that it works for your purposes, let me know and at least we can post that.

Thanks again for your post!

Andrew

[piskuliche]

Andrew, Thanks for the detailed reply! I’ve actually been chatting with David a bit already because I am working with gREMD of lipid bilayers which he worked on a bit. I actually already wrote a code that starts to read some of the MARTINI files to extract and build a moltemplate lt file for individual lipids, so hopefully it won’t be too hard to actually go the rest of the way and build an overall code like you are describing to actually build the rest of the force field file. I’ll work on it a bit and I will reach out to you again you once I have it! Thanks again for the response! Zeke

[jewettaij]

Cool! Glad to hear you are working with David. (not Daniel, arghh) I think the gREMD method is underappreciated. And I'm also glad to hear you are writing a converter. I would love to include that with moltemplate when it's ready, if that's okay with you. (Along with the lipid files.)
Thanks for posting. I will change the title of this post to "drymartini and martini issues".