Martini 3: a general purpose force field for coarse-grained molecular dynamics

Abstract

The coarse-grained Martini force field is widely used in biomolecular simulations. Here we present the refined model, Martini 3 ( http://cgmartini.nl ), with an improved interaction balance, new bead types and expanded ability to include specific interactions representing, for example, hydrogen bonding and electronic polarizability. The updated model allows more accurate predictions of molecular packing and interactions in general, which is exemplified with a vast and diverse set of applications, ranging from oil/water partitioning and miscibility data to complex molecular systems, involving protein-protein and protein-lipid interactions and material science applications as ionic liquids and aedamers.

Figure 1:. Rebalancing R-, S-, and T-beads –

(A) Snapshots of simulation boxes containing mixtures of dodecane and water in three resolutions. (B) Radial distribution functions (g_ij) for all bead combinations in the multi-resolution mixture of water. (C) Water-oil transfer free energies (ΔG) computed for around 260 data points using Martini 3. (D) Hydrogen bonding potential of mean force (PMF) between nucleobases. On the left, comparison between Martini 2 and 3 for the cytosine-guanine base pair. On the right, comparison of the cytosine-guanine (C-G) and guanine-guanine (G-G) base pairs using Martini 3. In both plots, CHARMM and AMBER atomistic data are also reported for comparison. Errors are estimated with bootstrapping and displayed as transparent shades. In the case of Martini, errors are smaller than 0.1 kJ/mol, and hence are not visible in the graphs.

Figure 2:. New chemical bead types, sub-labels, and applications –

(A) Self-assembly of aedamers. The left panel shows the dimerization free energies (ΔG_dim) of pegylated monomers of 1,5-dialkoxynaphthalene (DAN) and naphthalene diimide (NDI). Errors are estimated with bootstrapping. The right panel shows the self-assembled duplex dimer formed by amide-linked tetramers of NDI (green) and DAN (orange). (B) As indicatedby X-ray crystallography, sodium ions (charged TQ5 bead) can bind to a buried small cavity in the core of the adenosine A_2A receptor. (C) Charged Q-beads in Martini 3 follow the classical Hofmeister series, as exemplified by the anion transfer between salt aqueous solutions and organophosphonium-based ionic liquids (right panel). Errors in the average anion transfer percentage are estimated by block averaging. (D) Preferential cation-π interaction between choline groups (Q1 bead) of phosphatidylcholine lipids and aromatic residues of the Bacillus thuringiensis phosphatidylinositol-specific phospholipase C (BtPI-PLC). The depth of insertion of each amino acid of BtPI-PLC is in very good agreement with the insertion obtained from an atomistic MD simulation.

Figure 3:. Improving packing, cavities and reducing protein stickiness –

(A) Hydration of Fragaceatoxin C (FraC) nanopore inserted in a lipid bilayer. (B) Scattering profiles and a Martini 3 snapshot of a bulk heterojunction morphology of poly(3-hexyl-thiophene) (P3HT, in red) and phenyl-C61-butyric acid methyl ester (PCBM, in blue) formed after solvent evaporation and annealing simulations. I(q) corresponds to scattering intensity and q is the reciprocal space vector. (C) Aggregation levels of the soluble proteins villin headpiece and the modified EXG-CBM in different salt concentrations. (D) Aggregation levels of polyleucine helices in POPC and DLPC bilayers. Errors in the average monomer percentage of (C) and (D) are estimated by block averaging. (E) Dimerization of transmembrane helices. Left panel shows a comparison between experimental and calculated values for the mole fraction standard Gibbs free energy of dimerization $\left(\mathrm{\Delta }{G}_{ass}^X\right)$ of the following transmembrane protein domains: ErbB1, EphA1, WALP23 and GpA. Simulation errors are estimated with bootstrapping while experimental data was obtained in the literature^–. In the case of GpA, error was estimated by the mean absolute error of four independent experimental data^–. A comparison between experimental and simulated binding modes of GpA is highlighted in the right panel. The experimental structure was taken from solution NMR in micelles (PDB code AFO). Near identical experimental structures were obtained by ssNMR in nanodiscs and X-ray crystallography in a lipid cubic phase^–.