Abstract

Many critical process in biology span ranges of length and time scales that make infeasible their direct analysis by atomic-scale modeling. Instead, multiscale methods are essential to modeling, predicting, and understanding the basic physical driving forces that govern these systems. Here, we discuss two recent projects in our group that connect reduced-physics descriptions of biomolecular systems with detailed all-atom simulations of accurate physiochemical models. We first discuss efforts to understand and predict folding and mis-folding pathways of protein and peptide systems. We show that fragment-based approaches to the simulation of proteins help understand folding and aggregation pathways. We also describe new algorithms for probing the free energy surfaces underlying peptide dimerization. In the second part of the talk, we introduce a new, fundamental framework for multiscale studies based on an information-theoretic concept called the relative entropy. We discuss its use in the development of robust coarse-graining algorithms, and its application to interpreting liquid water’s important deviations from the behavior of simple liquids.