Modeling
Experiments and Experimental Observability
Various biophysical methods exist to probe the internal dynamics of molecules and the formation of molecular structures in cells. On both levels, experiments are limited in terms of observability: Ensemble molecular experiments, such as IR, NMR and neutron scattering can probe structural and kinetic aspects but only report on ensemble-averaged quantities that make it difficult to infer details about structure and dynamics. Single-molecule experiments such as single-molecule FRET and force probe experiments yield trajectories of single molecules, providing access to the conformational fluctuations. However, the complex conformational dynamics can still not be observed as only one or a few experimental observables can be tracked at once. Finally, microscopic experiments can track static and dynamic molecular assemblies, but are limited by optical resolution.
We have developed a general theoretical framework to optimally combine simulation data with kinetic experimental data of the same biological system. The keys to this are (1) computational methods which allow the complete probability distribution of Markov models corresponding to a set of simulation or experimental observations to be evaluated, and (2) Methods that allow the experimental features to be computed directly from the Markov model and unambiguously relate them to structures and structural changes via the Markov model Eigenvalues and Eigenvectors. This provides a unification of experiment and simulation that allows to explore the microscopic details of biomolecular transitions. We are pursuing to build Markov models directly from data of time-resolved single-molecule experiments in close collaboration with experimentalists. The theoretical challenge is to infer the relevant states and transition rates of the complex dynamical system from a low-dimensional observation time series. We develop inverse modeling methods based on Likelihood-based models such as Hidden Markov Models to deal with this. In previous work, we have constructed multi-state kinetic models of RNA and protein folding dynamics from single-molecule FRET experiments and optical tweezer force-probe experiments.
Recently, we have started to work on statistical methods that will improve the resolution of super-resolution microscopy.
For more information see markovmodel.org
Collaborators
Andres Jäschke | Univ. Heidelberg | Single-molecular FRET
Susan M. Marqusee | UC Berkeley | Force-probe experiments
G. Ulrich Nienhaus | KIT | Single-molecule FRET
Marcus Sauer | Univ. Würzburg | Fluorescence correlation spectroscopy