Information transfer within cells on the molecular level occurs via a mechanism called allostery, where binding of signaling molecules to a protein cause subsequent structural changes traveling through this protein to a distant site. The physical mechanism though which allostery takes place is under hot debate. Computational investigations of allosteric communication require the characterization of both structural and dynamical changes of the target protein via nonequilibrium molecular dynamics simulations.

Energy transport through molecular systems has recently received considerable interest, in particular due to its importance for molecular electronics and the functioning of biological systems. Here we employ nonequilibrium MD simulations, which mimic the laser excitation of the molecules by nonequilibrium phase-space initial conditions, and follow the flow of vibrational energy through a protein. To identify the pathways of the energy transport, we construct a master equation model, the transfer rates of which are obtained from two scaling rules, which account for the energy transport through the backbone and via tertiary contacts, respectively.

Biased MD simulations are an efficient tool to derive free energies along predefined reaction coordinates, but come at the cost of lost information on the dynamics of a molecular system. We have developed dissipation-corrected targeted molecular dynamics to not only extract free energies, but friction profiles along biasing coordinates, which allows the recovery of dynamics via Langevin equation framework.

The statistical analysis of MD simulations requires dimensionality reduction techniques, which yield a low-dimensional set of collective variables that in some sense describe the essential dynamics of the system. Those collective variables can be used to construct a free energy landscape which characterizes the low-energy regions and the energy barriers, i.e., the metastable states and transition regions of the system. We have developed various methods to identify collective variables and metastable states (e.g., using machine learning), which subsequently can be employed to construct a Langevin or a Markov state model of the dynamics.

Markov state models allow to approximate the dynamics of a protein by memory-less jumps between the system's metastable states. Focusing on the slowest processes of a system, they play a key role in the analysis and the enhanced sampling of MD data. Apart from pushing forward methods to construct Markov models, we are particulary interested in the modeling of nonequilibrium processes which are ubiquitous in biology.

The stochastic protein dynamics on a low-dimensional free energy landscape can be described by a Langevin approach. In particular, we have developed a data-driven Langevin equation formalism, which allows efficient estimation of multidimensional Langevin fields fom MD data. Recent research focuses on nonequilibrium processes and enhanced sampling schemes.