Interactions of Proteins with Interfaces and Solvents
Many novel materials show tremendous potential for helping to solve pressing challenges in creating a sustainable source of energy. Unfortunately, thermodynamic or kinetic limitations often prevent us from production at relevant scales, and thus limit the societal impact of many discoveries that first appear to be transformational in nature. Therefore, it is critical to develop the capability to predict and control the environment that hosts novel nanostructured materials in order to maximize widespread societal impact.
We use atomistic molecular simulations to study the behavior of biomolecules at interfaces; specifically, the interface between proteins and ionic liquids (ILs) with applications in biocatalysis and the interface between proteins and solid surfaces with applications in biomaterials, biomineralizaton, and biomass processing. For example, simulations have revealed the preferential binding patterns of IL solvents on enzyme active sites and suggest possible mechanisms of enzyme inhibition. We are also developing computational screening tools to help reduce experimental costs and time to develop new materials.
However, molecular simulation approaches such as equilibrium molecular dynamics (MD) are often insufficient for such examples as the accessible time and length scales accessible in the simulations do not correspond with scale over which interesting phenomena occur. We address this challenge by using and developing various techniques based on enhanced sampling with the metadynamics method or coarse grained modeling to accurately compute quantities such as binding free-energies of proteins at interfaces and rates of transition between low free-energy biomolecular states. In addition, we are studying the role in which the surface can change the favored conformational states of proteins, specifically looking at how features like surface roughness, charge, and headgroup functionality affect the free-energy landscape of biomolecules.
Analysis of Complex Reacting Systems
Reaction networks are fundamental for understanding complicated chemical and biological processes. Researchers have provided many methods for using computers to generate reaction networks using computational and mathematical approaches. While extremely powerful, these approaches require users to rely on their preconceptions about the chemistry and chemical intuition about what classes of reactions might occur. These preconceptions may lead to omission of important reactions or inclusion of superfluous reactions and impact the ultimate value of the mechanism.
We are using ab initio MD simulations combined with special non-equilibrium enhanced sampling methods to study many aspects of complex reacting systems. We seek to use MD-based approaches to discover new chemistry (i.e., reaction network topology) without relying on prior knowledge of reaction classes. For interesting classes of reactions such as enzyme or condensed phase, enhanced sampling methods can also be used to calculate the transition times and energy barriers between stable states in the reaction network. Finally, we seek to use these methods to study reactions with significant entropic components at finite temperatures.
Applications of Data Science in Chemical Engineering
From both the perspective of data intensive experiments and simulations, many researchers now face a common challenge: how to make best use of large data streams that are now routinely available. We apply the methods of Data Science to a variety of chemical engineering challenges. We are use artificial neural networks to enable process optimization and reduce the computation costs of running detailed kinetic models for various applications in clean energy. Decision trees and random forests are being applied to atomic force microscopy data to predict photoluminescence properties of new materials for energy storage and conversion. Design tools for new solvents and materials are being developed to discover new insights and relationships from large data sets based on massive campaigns of MD simulations.