(750g) A Hybrid Approach Toward Systematically-Derived Implicit-Solvent Coarse-Grained Lipid Models | AIChE

(750g) A Hybrid Approach Toward Systematically-Derived Implicit-Solvent Coarse-Grained Lipid Models


Pak, A. J. - Presenter, The University of Chicago
Dannenhoffer-Lafage, T., The University of Chicago
Madsen, J. J., The University of Chicago
Voth, G. A., The University of Chicago
Biological membranes are intimately involved in many cellular processes, including cell signaling, energy storage, compartmentalization, and active transport. These membranes are composed of a diverse array of phospholipids, which self-assemble into bilayers due to their amphiphilic nature. Many of the macroscopic properties of these membranes, and resultant insights into their biological functionality, would benefit from a molecular understanding of their behaviors. To this end, implicit-solvent coarse-grained (CG) models enable investigation into molecular dynamics at biologically-relevant length- and time-scales. However, as water is largely responsible for the entropy-driven hydrophobic effect that underlies lipid behavior, parameterization of effective CG interactions has been notoriously difficult. In this talk, we will present an approach to systematically derive high-fidelity, low-resolution CG lipid models from all-atom simulations by leveraging two techniques: multiscale coarse-graining (MSCG) and relative entropy minimization (REM). We first provide a comprehensive comparison between MSCG- and REM-derived CG lipid models, highlighting their respective benefits and deficiencies. Then, in contrast to conventional methodologies, we demonstrate the importance of featurizing interfacial water in the form of virtual CG particles; here, effective interactions are derived from a hybrid MSCG/REM framework. The resultant CG models recapitulate critical physical phenomena, including robust self-assembly with diverse morphologies and phases, which we attribute to a redistribution of entropic and enthalpic driving forces. As a final note, we discuss the potential to generalize our approach for biomolecular CG models with improved transferability.