Next-generation coarse-grained models for molecular dynamics simulations of fluid phase equilibria and protein biophysics using the relative entropy
Coarse grained models for molecular dynamic simulations of liquid structure and protein folding and self-assembly have been the subject of decades of research efforts. Although such models enable probing into longer length and time scales, they are still limited in accuracy and scale poorly to thermodynamic conditions or chemical entities beyond the ones at which they are developed. In this work, I leverage a powerful coarse graining framework based on an information theoretic metric known as the relative entropy to design novel coarse grained models for equilibrium phase behavior in liquid mixtures that correctly address the relevant manybody physics critically involved in phase behavior and thus, can provide structurally accurate descriptions of macroscopic phase separation across a large range of mixture compositions. I also develop protein models that are “next- generation” in the sense that they are systematically extendable in complexity while depending minimally on experimental data. These protein models offer remarkable predictive accuracy for folding of single proteins and insights into large scale protein self-assembly commonly seen in neurodegenerative pathologies such as Alzheimer’s and prion diseases. This dissertation develops and applies powerful coarse-graining techniques to meet longstanding challenges in the field and elevate coarse grained models from being “toy” systems to accurate, “production-level” tools for the development of biotechnologies and advanced materials.