The famous sampling problem consists of the fact that biomolecules can, due to the computational effort, not be simulated at biologically relevant timescales. We are developing methods that alleviate the famous sampling problem for a large class of molecular processes, including folding, binding and conformational changes, without destroying the dynamical information.
|
|
The approach exploits the fact that, for a given Markov model estimated from available data, the entire statistical distribution of models, including the uncertainties in all parameters, can be estimated as well. Based on these uncertainties, an optimization problem can be formulated whose result allows to plan new simulations in such a way that the uncertainties are reduced with minimal effort (Collaboration with J. Smith, IWR Heidelberg). This approach allows the sampling effort to be reduced by several orders of magnitude, such that MD simulations of slow biological processes, such as protein folding and protein:ligand binding, may be within reach [Noé et al. MMS 06, JCTC 06]. |
|
Besides the application on molecular dynamics, the optimal planning approach is also extended to a new fundamental problem in discrete mathematics, namely the daptive refinement of fuzzy graphs (Collaborations with G. Reinelt, IWR Heidelberg and M. Skutella, TU Berlin) [Noé, Oswald, Reinelt, OR 07]. People involved: Jan-Hendrik Prinz, Emal Alekozai
|
|
|
|
|
|
