Conformational dynamics of biomolecules: Reconciling simulation and experimental data

• Applicant: Professor Dr. Frank Noé
• Subject Area: Biophysics
• Term: from 2008 to 2019
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 104734058

• Final Report Year: 2021

Final Report Abstract

Biomolecules often possess multiple metastable states, e.g. protein(s) fluctuate within a set of structures, possibly associated with a particular biological function, for a long time before enough thermal energy is accumulated to leave this set and transition to another metastable set. It is the interest of chemical physicists and biophysicists to identify the essential metastable states, quantify their free energies or probabilities, the kinetics arising from the transitions between them, and the structural mechanisms involved. Several types of biophysical experiments are available to probe metastable states and the transitions between these metastable states, but there is a tradeoff between the sensitivity to dynamics and the ability to resolve structural detail. Mainly two types of experiments are considered here: (i) Ensemble kinetics experiments that time correlations of a complex spectroscopic signal, such as dynamical neutron scattering and NMR experiments can generate multidimensional information that depends on both structure and dynamics, but one cannot obtain detailed structures or which structural changes are responsible for particular relaxations or correlations observed. (ii) Single molecule experiments, such as single-molecule Förster resonance transfer (FRET), have a direct access to the dynamics of a process, such as opening/closing motions or folding/unfolding of a protein, but usually can only probe a single or a few structural features simultaneously. Molecular dynamics (MD) simulations are as yet the only technique which allow high-resolution structures and dynamics to be probed simultaneously. Despite continuous improvements in the quality of force-fields and increasing computational power and thus better sampling, MD has become quite reliable in predicting structures of stable states, such as the folded structure in fast folders or relevant structures of the binding pockets in receptors. However, due to their reliance on force-fields, i.e. parametric computational models of the protein, they may involve significant errors in quantitative prediction of the probabilities of conformations and the rates of their interconversion. In this project we have developed a suite of methods to combine experimental and simulation data in order to combine the best of both worlds, experiments and simulations. The key technology to “communicate” between the experimental and MD views are Markov States Models (MSMs), that have been pioneered and developed mainly by a few groups worldwide including the group of the PI. An MSM consists of (i) a subdivision of the state space into a discrete set of states derived usually from the simulation data by some combination of dimension reduction and clustering methods, or, more recently, also machine learning methods; and (ii) a Markovian model to describe the transition dynamics amongst these states, usually a transition probability matrix or rate matrix. In contrast to standard analyses of molecular dynamics simulations, (1) MSMs can predict long-term molecular kinetics from short-time simulations, (2) great amounts of simulation data can be analyzed with little subjectivity of the analyst, and (3) stationary and kinetic physicochemical quantities, such as conformational free energy differences, metastable states, and the ensemble of transition pathways can be easily calculated. MSMs are especially useful when studying complex macromolecular changes, such as folding, native-state conformational transitions, and binding. This project has developed a suite of MSM-related methods to combine experimental and simulation data in order to combine the best of both worlds: (i) the ability of MD simulations to predict detailed 3D structures and transition processes between long-lived states, (ii) the ability of experiments to probe the probabilities of these states and the transition rates between them quantitatively.

Publications

• Dynamical fingerprints for probing individual relaxation processes in biomolecular dynamics with simulations and kinetic experiments. Proc. Natl. Acad. Sci. USA 108, 4822-4827 (2011)
  F. Noé, S. Doose, I. Daidone, M. Löllmann, M. Sauer, J.D. Chodera and J.C. Smith
  
• Markov models of molecular kinetics: Generation and validation. J. Chem. Phys. 134, 174105 (2011)
  J.-H. Prinz, H. Wu, M. Sarich, B. Keller, M. Senne, M. Held, J.D. Chodera, C. Schütte and F. Noé
  
• Markov models and dynamical fingerprints: Unraveling the complexity of molecular kinetics. Chem. Phys. 396, 92-107 (2012)
  B.G. Keller, J.-H. Prinz, F. Noé
  
• Dynamic neutron scattering from conformational dynamics. I. Theory and Markov models J. Chem. Phys. 139, 175101 (2013)
  Benjamin Lindner, Zheng Yi, Jan-Hendrik Prinz, Jeremy C. Smith, and Frank Noé
  
• Dynamic neutron scattering from conformational dynamics. II. Application using molecular dynamics simulation and Markov modeling J. Chem. Phys. 139, 175102 (2013)
  Zheng Yi, Benjamin Lindner, Jan-Hendrik Prinz, Frank Noé, and Jeremy C. Smith
  
• Projected and hidden Markov models for calculating kinetics and metastable states of complex molecules J. Chem. Phys. 139, 184114 (2013)
  Frank Noé, Hao Wu, Jan-Hendrik Prinz, and Nuria Plattner
  
• Complex RNA Folding Kinetics Revealed by Single-Molecule FRET and Hidden Markov Models. J. Am. Chem. Soc. 136, 12, 4534–4543 (2014)
  Bettina G. Keller, Andrei Kobitski, Andres Jäschke, G. Ulrich Nienhaus and Frank Noé
  
• Optimal estimation of free energies and stationary densities from multiple biased simulations, Multiscale Model. Simul. 12, 25–54 (2014)
  Hao Wu and Frank Noé
  
• Spectral Rate Theory for Two-State Kinetics, Phys. Rev. X 4, 011020 (2014)
  Jan-Hendrik Prinz, John D. Chodera and Frank Noé
  
• Combining experimental and simulation data of molecular processes via augmented Markov models. Proc. Natl. Acad. Sci. USA 114:8265-8270. (2017)
  Simon Olsson, Hao Wu, Fabian Paul, Cecilia Clementi and Frank Noé
  
• Mechanistic Models of Chemical Exchange Induced Relaxation in Protein NMR. J. Am. Chem. Soc. 139, 200–210 (2017)
  Simon Olsson and Frank Noé
  
• OpenPathSampling: A Python Framework for Path Sampling Simulations. 1. Basics. J. Chem. Theory Comput. 15, 813–836 (2019)
  David W. H. Swenson, Jan-Hendrik Prinz, Frank Noe, John D. Chodera, and Peter G. Bolhuis