Kinetics of molecular systems occur often on microsecond or higher timescales. Thus, investigation and sufficient sampling of such systems with computer simulations like molecular dynamics (MD) is challenging. External biasing potentials can accelerate such slow kinetics, but at the expense of altering the kinetic and thermodynamic properties of the system. Only recently, dynamic reweighting algorithms have been proposed to recover the unbiased kinetics (and populations) from biased systems [1-3]. However, especially the highly accurate path reweighting algorithms suffer from various implementation problems, i.e., the reweighting has to be performed on-the-fly, is integrator dependent, and only applicable to stochastic dynamics (SD).

We introduce a new, user-friendly path reweighting algorithm that can be used with any numerical scheme. Hence, our approach extends the scope of path reweighting methods to deterministic MD and is independent of the integrator used. Additionally, the new formulation leads to an increased parameter stability. Our path reweighting algorithm brings us thus closer to the goal of an easily applicable and robust reweighting scheme, which will be crucial to study and understand the dynamics of increasingly complex systems.

[1] Hao Wu, Fabian Paul, Christoph Wehmeyer and Frank Noé, Proc. Natl. Acad. Sci. U.S.A., 2016, 113, 3221-3230
[2] Lukas S. Stelzl, Adam Kells, Edina Rosta, and Gerhard Hummer, J. Chem. Theory Comput., 2017, 13, 6328–6342
[3] Luca Donati and Bettina Keller, J. Chem. Phys., 2017, 149, 072335