Pitfalls of the Martini Model

Abstract

<div> <div> <div> <p>The computational and conceptual simplifications realized by coarse-grain (CG) models make them an ubiquitous tool in the current computational modeling landscape. Building block based CG models, such as the Martini model, possess the key advantage of allowing for a broad range of applications without the need to reparametrize the force field each time. However, there are certain inherent limitations to this approach, which we investigate in detail in this work. We first study the consequences of the absence of specific cross Lennard-Jones parameters between different particle sizes. We show that this lack may lead to artificially high free energy barriers in dimerization profiles. We then look at the effect of deviating too far from the standard bonded parameters, both in terms of solute partitioning behavior and solvent properties. Moreover, we show that too weak bonded force constants entail the risk of artificially inducing clustering, which has to be taken into account when designing elastic network models for proteins. These results have implications for the current use of the Martini CG model and provide clear directions for the reparametrization of the Martini model. Moreover, our findings are generally relevant for the parametrization of any other building block based force field. </p> </div> </div> </div>

Protein–ligand binding with the coarse-grained Martini model

“…The examples showed in this work indicate that the current state of the Martini model seems to finally achieve most of these requirements with reasonable accuracy in relation to atomistic models. The key improvement in Martini 3 33 to enable such applications is the enhanced packing of the CG beads, achieved by re-balancing of the cross-interactions of different bead sizes 78 , as well as by the re-parametrization of bonded distances based on molecular volume and shape. As a result, protein cavities are represented more realistically and ligands can fit better.…”

“…The bonded parameters of the protein models were slightly adapted from the standard Martini 2.2 settings including the recently suggested side-chain corrections 93 , applied not only for β-strands but to all loops and secondary structure elements. An elastic network comparable to the one of the Martini 2.2 protein model 88 was used to maintain the secondary and ternary protein structure without exclusions of the non-bonded interactions between the backbone beads connected by the elastic network 78 . See more details about the protein models in the Supplementary Methods .…”

Modulation of A2aR Oligomerisation by Conformational State and PIP2Interactions Revealed by MD Simulations and Markov Models

“…A number of simulation studies using the MARTINI force field have provided valuable insights into the association of membrane proteins 28,32,[49][50][51][52][53] and into protein-lipid interactions [54][55][56] . However, the current MARTINI model may also have some limitations 57,58 . In particular, the lack of size-specific In the current study we have employed relatively simple models of the lipid mixture present in mammalian plasma membranes, as in our previous studies of GPCRs 35,59 .…”

Membrane mediated toppling mechanism of the folate energy coupling factor transporter

“…Free energy calculation. As protein-protein interactions in Martini 2 are usually overestimated 45,46 , we resorted to the recently reparametrized version of the Martini force field, called Martini 3 47,48 , to compute the unbinding free energy profile of the S-component from the ECF complex. CG protein models were generated using the new version of the program Martinize 36,49 .…”