I have been interested in using MSM to extend the capability of MD simulations. The fundamental question I am interested in answering if whether we can reduce the computational complexity of a new simulation by leveraging information from the current MSM. The new simulation could be a new sequence, a new physics model or perhaps some post translational modification. To that end, I am borrowing ideas from Reinforcement-Learning and Markov decision processes.

Recently, I showed over a series of papers that the tICA algorithm can be used to form better CV than human intuition in enhanced sampling algorithms such as Metadynamics/Umbrella sampling. This is