Enhancing the Assembly Properties of Bottom-Up Coarse-Grained\nPhospholipids

A plethora of key\nbiological events occur at the cellular membrane\nwhere the large spatiotemporal scales necessitate dimensionality reduction\nor coarse-graining approaches over conventional all-atom molecular\ndynamics simulation. Constructing coarse-grained descriptions of membranes\nsystematically from statistical mechanical principles has largely\nremained challenging due to the necessity of capturing amphipathic\nself-assembling behavior in coarse-grained models. We show that bottom-up\ncoarse-grained lipid models can possess metastable morphological behavior\nand that this potential metastability has ramifications for accurate\ndevelopment and training. We in turn develop a training algorithm\nwhich evades metastability issues by linking model training to self-assembling\nbehavior, and demonstrate its robustness via construction of solvent-free\ncoarse-grained models of various phospholipid membranes, including\nlipid species such as phosphatidylcholines, phosphatidylserines, sphingolipids,\nand cholesterol. The resulting coarse-grained lipid models are orders\nof magnitude faster than their atomistic counterparts while retaining\nstructural fidelity and constitute a promising direction for the development\nof coarse-grained models of realistic cell membranes.