Room for improvement in the initial martini 3 parameterization of peptide interactions

The Martini 3 coarse-grain force field has greatly improved upon its predecessor, having already been successfully employed in several applications. Here, we gauge the accuracy of Martini 2 and 3 protein interactions in two types of systems: coiled coil peptide dimers in water and transmembrane peptides. Coiled coil dimers form incorrectly under Martini 2 and not at all under Martini 3. With transmembrane peptides, Martini 3 represents better the membrane thickness–peptide tilt relationship, but shorter peptides do not remain transmembranar. We discuss related observations, and describe mitigation strategies involving either scaling interactions or restraining the system. [Display omitted] •Martini 3 coarse-grain fails to capture coiled-coil helix dimerization.•Helix contacts are sparse and occur via wrong interfaces.•Martini 3 recovers transmembrane helix tilt dependency on thickness mismatch.•A possible hydrophobicity imbalance drives out helices from the membrane core.