One of the key limitations of Molecular Dynamics (MD) simulations is the computational intractability of sampling protein conformational landscapes associated with either large system size or long time scales. To overcome this bottleneck, we present the REinforcement learning based Adaptive samPling (REAP) algorithm that aims to efficiently sample conformational space by learning the relative importance of each order parameter as it samples the landscape. To achieve this, the algorithm uses concepts from the field of reinforcement learning, a subset of machine learning, which rewards sampling along important degrees of freedom and disregards others that do not facilitate exploration or exploitation.
In this package, we present a demo of REAP and provide the original source code used for the testing the performance of the algorithm in the publication; "Shamsi, Z., Cheng, K. J., & Shukla, D. (2018). Reinforcement learning based adaptive sampling: REAPing rewards by exploring protein conformational landscapes. The Journal of Physical Chemistry B".
Python code to use REAP algorithm to find the best starting points for each round of simulations.
Dependencies:
- numpy
- msmbuilder
- mdtraj
For a tutorial on how to use REAP on MD simulations and toy potentials to get starting structures for the next simulation round, please see the included file Tutorial-LShapedLandscape.ipynb.
If you use this code, please cite the following paper:
Shamsi, Z., Cheng, K. J., & Shukla, D. (2018). Reinforcement learning based adaptive sampling: REAPing rewards by exploring protein conformational landscapes. The Journal of Physical Chemistry B, 122(35), 8386-8395.