Coarse-graining PAMA polymers with polyply, AutoMartini-M3, bartender, and martinize2

[jotelha]

Dear MARTINI community,

I have recently started on a PostDoc-Position at Nagoya University. In my project, I am tasked to develop CG models for PAMA (poly(alkyl methacrylate)) polymers, such as those described in

H. Shekhar, Y. Chikazawa, Y. Song, and H. Zhang, “Bulk conformation and shear-induced transitions in bottlebrush poly(alkyl methacrylate): The role of side chain length and polymer flexibility,” Tribology International, vol. 204, p. 110476, Apr. 2025, doi: 10.1016/j.triboint.2024.110476.

The ultimate goal is to run shearing simulations on fully-developed adsorption films of such polymers to investigate the relation between polymer design parameters and lubricating performance.

An important aspect of the project is to make the model development robust and reproducible, ideally casting the coarse-graining procedure into automatable workflows without much human interference.

I have reviewed Martini-related literature and found that with the latest Martini 3 iteration, the tooling puts an emphasis on enabling high throughput screening. Hence, I would like anchor my coarse-graining project in the Martini 3 framework and make as much use of the provided tooling and best practices as possible.

In the following, I describe what I have done so far and what I envision, and I would be happy about second opinions on my approach. I have largely followed a tutorial at https://vermouth-martinize.readthedocs.io/en/v0.11.0/tutorials/6_adding_residues_links/index.html .

The minimal set of monomers I am dealing with are listed in the following table. I might want to explore other monomers later on, hence the strong focus on automizability and repeatability of the CG workflow.

	full name	formula	Martini 3 beads
EH	2-Ethylhexyl methacrylate	C12H22O2	SC4-TN5a-TP2a-SC3-TC3-TC3
LA	Lauryl methacrylate	C16H30O2	SC4-TN5a-TP2a-TC3-SC3-TC3-TC3-TC3
MMA	Methyl methacrylate	C5H8O2	SC4-P2a
OC	Octyl methacrylate	C12H22O2	SC4-TN5a-TP2a-TC3-SC3-TC3
ST	Stearyl methacrylate	C22H42O2	SC4-TN5a-TP2a-TC3-TC3-SC3-TC3-TC3-TC3-TC3-TC3

I have retrieved structures for all these monomers from the ATB server and converted them to SDF. Next, I have fed these SDF into AutoMartini-M3 to programmatically assign Martini 3 coarse-graining beads. The generated bead types are also shown in the table above.

Next, I have used cgbuilder to manually map the bead assignment back onto the original atom names as contained in my ATB-generated structures. Unfortunately, the pathway of SDF files fed into AutoMartini-M3 does not preserve atom names. I would be grateful for a bead assignment procedure that preserves atom names.

After bead assignment, I have used bartender and xtb with the AutoMartini-M3-generated input files to optimize bonded interactions between beads for each of these monomers.

Subsequently, I have used martinize2 to to generate CG representations from PAMA AA models. For this purpose, I have created a custom vermouth data library extension, https://github.com/jotelha/vermouth-pama, using the ATB-generated itp files for the AA force field library entries, the AutoMartini-M3-generated itp files for CG force field library entries, and the cgbuilder-generated mapping files for mapping library entries.

Simulataneously, I have started experimenting with polyply for generating arbitrary AA PAMA from generic polymerization descriptions in string format. For this purpose, I have forked polyply and added custom OPLS-AA force field files for my monomers generated with the LibParGen server, https://github.com/jotelha/polyply_1.0/tree/2025-12-01-PAMA-parameters.

I am now at the point, where I can generate sequences of monomers and their topologies with commands like

polyply gen_seq -from_string A:54:1:MMA-1.0 B:73:1:EH-0.452,LA-0.288,ST-0.260 -o 1.json -name PAMA_BLOCK -seq B A

polyply gen_params -lib oplsaaLigParGen -o 1.itp -name PAMA_BLOCK -seqf 1.json

And I can also map AA representations to their respective CG counterparts with commands like

martinize2 -ff-dir force_fields -map-dir mappings -f default.pdb -x default_CG.pdb -o topol.top -dssp -maxwarn 1

Now comes the step where I would appreciate some advice.

Both in the OPLS-AA model and in the Martini CG model, I now have to define "links" between monomers in their vermouth library representation. For this, I have to pick correct bonded and angular parameters. Here, I believe the automatability ends and the manual fine tuning begins.

Do you have recommendations on how to proceed picking these "backbone" interactions?

My naive approach would be as follows:

For the AA level, pick standard OPLS-AA parameters for the atom types involved in backbone bonds and angles.

For the CG level, I favor parameters introduced by

G. Campos-Villalobos, F. R. Siperstein, and A. Patti, “Transferable coarse-grained MARTINI model for methacrylate-based copolymers,” Mol. Syst. Des. Eng., vol. 4, no. 1, pp. 186–198, Feb. 2019, doi: 10.1039/C8ME00064F.

in their Table 5, specifically

Bonded interaction	$b_0 (nm)$, $\theta_0 (deg)$	$k_b (kJ mol^{-1} nm^{-2})$, $k_\theta (kJ mol^{-1})$
MB-MB bond	0.289	21100
MB-MB-MB angle	175	13

for backbone beads of Martini representations of PMMA as initial values. I would then likely run AA and CG simulations of single polymers in hexadecane and compare their radii of gyration, an possibly also simulations of polymer melts to compare their densities, and tune the CG backbone parameters accordingly if necessary.

What do you think of this approach? Will it be sufficient to use just one bond and angle type along the CG backbone, or will I have to differentiate between the 15 possible bond combinations between the five monomers EH, LA, MMA, OC, and ST, and the even higher number of angular interactions? And for the backbone interactions at the AA level, are there any recommended practices for writing links for OPLS-AA-described polymers?

Thank you very much for your advice.

Best,

Johannes