Extending the MARTINI 3 Coarse-Grained Forcefield to Polypeptoids

This study presents the first MARTINI 3-compatible coarse-grained forcefield for 19 common peptoid residues, derived from all-atom reference simulations and integrated into the martinize2 tool, to enable efficient and accurate mesoscale modeling of peptoid structures, self-assembly, and interactions.

The Gist

Imagine you are trying to understand how a complex machine works. You have two choices: you can look at every single screw, gear, and wire (the "All-Atom" view), or you can look at the machine as a collection of larger, functional blocks (the "Coarse-Grained" view).

This paper is about building a new, better set of "functional blocks" for a specific type of synthetic material called Polypeptoids.

Here is the story of what the researchers did, explained simply:

1. The Problem: The "Slow-Motion" Material

Polypeptoids are like artificial proteins. They are synthetic chains used to build smart materials, drug delivery systems, and tiny nanowires.

• The Catch: Unlike natural proteins, polypeptoids are very flexible and twisty. They have a special "hinge" in their backbone that can flip back and forth (like a door swinging open and closed).
• The Simulation Nightmare: To study these materials on a computer, scientists usually use "All-Atom" models. This is like trying to watch a movie where every single frame is drawn by hand. It's incredibly accurate, but it takes forever. Because the polypeptoid hinges flip so slowly, a computer simulation might run for weeks just to see the molecule twist once. It's like trying to watch a snail race by counting every step it takes.

2. The Solution: The "MARTINI 3" Toolkit

Scientists have a popular toolkit called MARTINI that simplifies molecules. Instead of drawing every atom, it groups them into "beads" (like LEGO bricks). This makes simulations run thousands of times faster.

• The Gap: Until now, there was no good "LEGO set" for polypeptoids that worked with the newest version of the toolkit (MARTINI 3). Existing sets were either too clunky or didn't fit with other materials (like cell membranes) that scientists wanted to study alongside polypeptoids.

3. The Breakthrough: Building the New LEGO Set

The team at the University of Alabama built the first-ever MARTINI 3-compatible model for polypeptoids. Here is how they did it:

• The Blueprint (Mapping): They had to decide how to turn a complex polypeptoid molecule into simple beads. They tested different ways to group the atoms. Imagine trying to pack a suitcase: you can fold clothes tightly (Mapping 1) or stuff them in loosely (Mapping 2). They found the "tight fold" that kept the shape of the molecule accurate without losing its flexibility.
• The Training (Learning from the Slow-Motion): To make sure their new "fast" model was accurate, they first ran the super-slow, super-accurate "All-Atom" simulations. They used a special trick called Parallel Bias Metadynamics (think of it as a helpful guide pushing the molecule through its slow twists so they could see all the possible shapes it could take).
• The Translation (Boltzmann Inversion): They took the data from those slow simulations and mathematically "translated" it into rules for their new fast model. This ensured that when the fast model moved, it behaved exactly like the slow, accurate one.

4. The Results: Speed vs. Accuracy

They tested their new model against the "gold standard" (the slow simulations) in four ways:

1. Size: Does the chain curl up or stretch out correctly? Yes.
2. Twisting: Does the backbone flip correctly? Yes.
3. Sticking: Do two chains stick together the right way? Yes.
4. Building: Do they form the right shapes when mixed together? Yes.

The Magic Number: The new model is 57 times faster than the old way.

• Analogy: If the old way took 57 hours to simulate a day of molecular activity, the new way takes just 1 hour. This allows scientists to watch these materials assemble and interact in real-time, rather than waiting months for results.

5. Why This Matters

This isn't just a speed upgrade; it's a key to the future.

• Rational Design: Scientists can now design new polypeptoid materials on a computer, test them instantly, and predict how they will behave before ever mixing chemicals in a lab.
• Universal Language: Because this model fits into the standard MARTINI 3 ecosystem, you can now easily mix polypeptoids with cell membranes, drugs, or other proteins in a single simulation. It's like speaking a universal language that allows different types of molecules to "talk" to each other in the computer.

Summary

The researchers took a slow, difficult-to-study material (polypeptoids), created a simplified but highly accurate "LEGO" version of it, and made it compatible with the world's most popular simulation toolkit. The result is a tool that is 57 times faster, allowing scientists to design the next generation of life-saving drugs and smart materials much more quickly.

Technical Summary

1. Problem Statement

Polypeptoids (poly-N-substituted glycines) are synthetic peptidomimetics where side chains are attached to the backbone amide nitrogen rather than the $\alpha$-carbon. This structural difference eliminates backbone hydrogen bonding, resulting in unique conformational flexibility and slow cis/trans $\omega$-dihedral isomerization. While all-atom (AA) force fields exist (e.g., CHARMM-compatible MoSiC-CGenFF-NTOID), they are computationally expensive and limited to small systems and short timescales (sub-microseconds).

Existing coarse-grained (CG) models for peptoids are largely confined to the MARTINI 2 framework or bespoke designs that lack compatibility with modern biomolecular ecosystems. The MARTINI 3 force field offers improved interaction schemes and bead definitions but has not yet been extended to polypeptoids. Developing a MARTINI 3-compatible model is critical to enable mesoscale simulations of peptoid self-assembly, membrane interactions, and large-scale nanostructure formation, which are currently inaccessible to AA simulations.

2. Methodology

The authors developed a bottom-up CG force field using a rigorous parameterization workflow:

• Reference Simulations (All-Atom):
  • System: Homopeptoid chains (4 to 20 residues) capped with Acetyl and Aminomethyl groups.
  • Force Field: MoSiC-CGenFF-NTOID (CHARMM-compatible).
  • Enhanced Sampling: Parallel Bias Metadynamics (PBMetaD) was employed to overcome the high activation barrier (~63 kJ/mol) of the backbone $\omega$-dihedral cis/trans isomerization. This ensured converged sampling of the conformational space, which is difficult to achieve with standard MD.
• Mapping Strategy:
  • Adopted a Center-of-Geometry (COG) mapping scheme (standard for MARTINI 3) rather than Center-of-Mass to preserve molecular volume.
  • Backbone: Evaluated four mapping schemes. Mapping 1 was selected, placing the geometric center along the C–C$\alpha$ bond using only heavy atoms (N, C$\alpha$, C, O). This yielded a unimodal, symmetric bond length distribution suitable for harmonic potentials.
  • Side Chains: Mapped to standard MARTINI 3 bead types (P2 for backbone, various classes for side chains) without introducing new bead types, ensuring compatibility with the existing chemical ontology.
• Parameterization:
  • Bonded Interactions: Derived via Direct Boltzmann Inversion (DBI) of probability distributions from PBMetaD trajectories.
    • Unified Approach: Most backbone terms (bonds, angles, dihedrals) were averaged across residues to create a unified backbone potential compatible with MARTINI 3's unified representation.
    • Clustered Approach: The BB–BB$^+$–SC1$^+$ angle showed significant residue dependence due to cis/trans constraints. The authors used k-means clustering to group 18 residues into 5 distinct clusters, assigning specific parameters to each.
    • Special Cases: Residues with charged side chains directly attached to the backbone (e.g., NAsp, Nae) required specific improper dihedral refinements to prevent conformational collapse.
  • Nonbonded Interactions: Primarily adopted from the standard MARTINI 3 library, leveraging the rebalanced interaction matrix to avoid excessive aggregation.
• Integration: The workflow was integrated into the martinize2 tool, allowing for automated topology generation.

3. Key Contributions

1. First MARTINI 3 Peptoid Force Field: Developed a transferable CG model covering 19 commonly used residue types, fully compatible with the modern MARTINI 3 framework.
2. Robust Sampling of Rare Events: Utilized PBMetaD to generate converged AA reference data for the slow cis/trans $\omega$-dihedral transitions, a defining feature of peptoid flexibility.
3. Unified yet Flexible Parameterization: Created a unified backbone parameter set for most interactions while employing a clustering strategy for residue-specific angle variations, balancing transferability with chemical accuracy.
4. Automated Workflow: Integrated the model into the martinize2 tool and released it via GitHub, enabling the research community to easily generate MARTINI 3 peptoid structures and topologies.
5. Performance Gain: Achieved a ~57-fold increase in computational efficiency compared to all-atom simulations while maintaining structural and thermodynamic fidelity.

4. Results and Validation

The model was rigorously validated against AA reference data across multiple observables:

• Polymer Scaling (Radius of Gyration, $R_g$):
  • The CG model reproduced AA $R_g$ trends for short chains ($N \le 10$) with <30% deviation for most residues.
  • Exception: NAsp (acidic) showed significant over-compaction in CG due to charge localization on a single bead. This was corrected by introducing residue-specific improper dihedrals (NAsp*), improving agreement for short chains.
• Backbone Conformational Landscape:
  • Free Energy Surfaces (FES) of backbone dihedrals ($\psi_1, \psi_2$) from CG simulations matched the broad, converged landscapes obtained from PBMetaD-enhanced AA simulations.
  • The model successfully captured the accessible conformational space, including regions restricted in unbiased AA simulations.
• Dimerization Thermodynamics:
  • Potentials of Mean Force (PMF) for dimerization of hydrophobic (NVal), polar (NCys), cationic (NArg), and aromatic (NPhe) tetramers showed excellent agreement with AA data.
  • Dimerization free energies differed by less than 1.69 $k_BT$ across systems, indicating accurate capture of residue-specific association behaviors.
• Sequence-Dependent Aggregation:
  • Simulations of 50-chain heterotrimer systems (FXF, FKF, KFF, XFF) successfully reproduced experimental trends: FXF (phenyl-lysine-phenyl analog) formed linear nanowires, while others formed globular structures.
  • While the CG model slightly overestimated cluster sizes (a known MARTINI tendency), it correctly captured the sequence-dependent aggregation hierarchy and orientational ordering.

5. Significance

This work establishes a transferable and computationally efficient framework for simulating polypeptoids at the mesoscale. By bridging the gap between atomistic detail and large-scale dynamics, the new force field enables:

• Rational Design: Facilitating the design of next-generation sequence-specific functional materials, drug delivery systems, and biomimetic nanostructures.
• Scalability: Allowing researchers to simulate large-scale peptoid assemblies, membrane interactions, and self-assembly processes that were previously computationally prohibitive.
• Ecosystem Integration: Extending the versatility of the MARTINI 3 ecosystem to include synthetic foldamers and sequence-defined polymers, paving the way for future developments in multiscale modeling of non-natural biopolymers.