Bruce Berne of Columbia University is supported by an award from the Theoretical and Computational Chemistry program to pursue theoretical and computational research with a focus on (a) large scale hydrophobicity and its mediation by cosolutes; (b) the role of force in protein unfolding and refolding and in mediating enzymatic reactions ; (c) the development of new efficient methods for sampling rough energy landscapes and new molecular dynamics (MD) methods that build on the multiple time scale methodologies developed in the PI's research group. This research addresses some problems that have long resisted explanation such as an explanation of the Hofmeister series and the mechanism by which urea acts as a denaturant. Topic (b) focuses on molecular mechanisms of protein folding and enzyme reactions under applied force as experimentally studied in force microscopy (AFM).
The research has broad applicability in many scientific areas including chemistry, biology, nanoscience and engineering
Liquids, solutions, glasses, membranes, and their interfaces are complicated disordered systems whose molecules are in continuous interaction with each other. Computer simulations have made it possible to un- derstand the structure and dynamics of such systems in a way that was not possible using more traditional theoretical approaches, and because of this one of the important aims of my research is the development of new computational algorithms for simulating these systems. In molecular dynamics (MD) the computer is used to solve Newton’s equations of motion for the atoms and molecules in the liquid and thus to generate a list of their positions and velocities as time evolves. This list provides all of the information needed to predict how the properties of the liquid evolve in time and it can be used to determine transport proper- ties, like viscosity, thermal conductivity, and reaction rates of chemical reactions, as well as the evolution the structure of the and thermodynamic properties of the system. Another approach is to use the Monte Carlo (MC) method, a method based on playing a certain kind of game of chance, to sample the positions of and conformations of molecules in an equilibrium system but this method cannot be used to calculate the dynamic or transport properties. The earliest formulations of MD or MC, although correct, often turn out to be inefficient when used to simulate complex systems of interest in material science or biology. Such simulations might take too much time even on the most advanced computers to be useful. Thus it is important to invent new approaches to make such simulations a useful tool. During this grant period we invented important new approaches to both MD and MC methodology. One of the causes of inefficiency in MD arises from very high frequency intramolecular vibrations of hy- drogen atoms in the constituent molecules, especially;y in OH bonds. The problem is important because it forces one to solve the equations of motion for very small time steps and thus requires one to perform a huge number of iterations in order to observe the relatively slow structural changes of interest say in condensed or biological systems. For example, to observe protein folding might require more than a trillion iterations for a system containing on the order of 200,000 atoms. This requires enormous computer resources. To tame this problem we devised a new method to damp out high frequency motions in molecular dynamics and greatly decrease the number of iterations required. Technically we accomplish this by introducing a "colored noise" thermostat. We showed that the use of this device allows us to reduce the effect of the fast vibrations without altering the important slow structural relaxations that we want to monitor. Likewise MC simulations of biological systems often involve a very large molecule (such as a protein) dissolved in a very large number of water molecules. Such simulations are time consuming because much time is wasted sampling new positions of water molecules far away from the region of interest such as the active site of a protein. We devised a significantly more efficient new method called REST (an acronym for Replica exchange with solute tempering) which allows us to greatly speed up this sampling, and we combined this with another method. This allowed us to be to accurately determine how the binding affinity of drug molecules (ligands) to proteins changes when chemical groups are added or removed from drug molecule, a problem of central importance to computer aided design of new drugs. The main tools used in my research are theory and simulation. As in the past, we developed new theoretical and computational methodologies including efficient methods for sampling rough energy land- scapes and new MD methods building on our multiple time scale methodologies. Our work during the tenure of this grant resulted in twenty one publications.1–21 In addition to the obvious benefits of the re- search this grant was central in training several Ph.D. students and postdoctoral research personnel. Several commercial software products produced by private companies are based in part on the research funded by this research. During the tenure of our NSF grant we used these new methods and some older ones to study the following questions: how quantum mechanical motions affect glasses; how temperature and the addition of osmolytes and anti-freeze proteins affect the nucleation of ice and the kinetics of freezing; how large-scale hydrophobicity affects the chemical kinetics of ligand-protein binding; and how quantum mechanics gives rise to isotope fractionation.
