Faculty of Engineering and Natural Sciences
Algorithms for Analyzing Molecular Motion and NMR Structure-Based Assignments
Mehmet Serkan Apaydın, Duke University
Classic techniques to simulate molecular motion, such as molecular dynamics (MD) or Monte Carlo (MC) simulation, generate individual pathways and spend most of their time in the local minima of the energy landscape defined over a molecular conformation space. Due to their high computational cost, it is impractical to compute ensemble properties, that is, properties requiring the analysis of many molecular pathways, using such techniques. We recently introduced Stochastic Roadmap Simulation (SRS) as a new computational framework for exploring the kinetics of molecular motion by simultaneously examining many pathways. These pathways are compactly encoded in a graph, which is constructed by sampling a molecular conformation space at random. Each arc in the graph represents a potential transition of the molecule and is associated with a probability indicating the likelihood of this transition. By viewing the graph as a Markov chain, we compute ensemble properties efficiently. This computation does not trace any particular pathway explicitly and circumvents the local minima problem. Furthermore, we formally show that SRS converges to the same stationary distribution as MC simulation.
We use SRS to study both protein folding and ligand-protein binding. In the former application, we measure the “kinetic distance” of a protein's conformation from its native state with respect to its unfolded state, using an important parameter, called probability of folding (pfold). We compare our pfold computations to those from MC simulation on a two-dimensional fictitious energy landscape, as well as for three proteins with different representations and energy functions. We find that SRS produces accurate results, while reducing the computation time by several orders of magnitude. We then estimate the transition state ensemble using SRS and predict folding rates and phi values for 16 proteins. Comparison with experimental data shows that SRS estimates the TSE much more accurately than an existing method based on dynamic programming. This leads to better folding-rate predictions. The results on phi value predictions are mixed, possibly due to the simple energy model used in the tests. In the latter application of SRS, we estimate the expected time to escape from a protein binding site for a ligand. Similar to pfold, it would be impractical to compute the escape time from a binding site with MD or MC simulations. We use escape time to qualitatively analyze the role of amino acids in the catalytic site of an enzyme by computational mutagenesis. These applications establish SRS as a new approach to efficiently and accurately compute ensemble properties of molecular motion.
We also discuss our recent work on using an ensemble of structures to compute NMR protein assignments.
Short biography: Mehmet Serkan Apaydın is a post-doctoral researcher in Prof. Bruce R. Donald's group at Duke University. He completed his undergraduate studies at Bilkent University (1997) and received his M.S. (1999) and Ph.D. degrees (2004) from Stanford University, where he was a member of the Stanford Robotics Lab working with Prof. Jean-Claude Latombe and Prof. Doug Brutlag. His research interests include NMR protein structure determination, molecular flexibility, protein folding, ligand-protein binding and motion planning. He is a recipient of David L. Cheriton Stanford graduate student fellowship and a bronze medalist in the 5th International Olympiad of Informatics (1993).
December 11, 2007, 14:40, FENS G029