Supplementary material for 'Towards a structural identification of metastable molecular conformations'

The generated simulation trajectories are given and some processed data: The end-to-end distances for the produced trajectories which are needed for the calculation of free energies. Jupyter notebooks are given which use the raw data and follow the modeling pipeline to the Markov state modeling step.

Latent spaces including training EncoderMap can be reproduced this way as well as the respective MSMs. Furthermore two interactive Jupyter notebooks in which the free energy landscape of both of the investigated dimensionality reduction methods TICA and EncoderMap with its respective average structures can be explored. Scripts for setting-up the environment to run these Jupyter-notebooks are provided.