Clustering the MSM

For now in BestMSM there is only one approach for clustering the microscopic MSM into a macrostate model. Here we borrow the methods described by Buchete and Hummer to build macrostates based on the sign or distribution of the eigenvector values.

• Coarse Master Equations for Peptide Folding Dynamics J. Phys. Chem. B 112 (19) 6057–6069 (2008).

As they stated in their paper, this approach is a version of Perron clustering (i.e. PCCA). On top of the methods described in the reference above, and following work by Chodera et al we include an optimization step in the clustering. This way, there is some chance that imperfections in the initial, eigenvector-based clustering, will be overcome. The optimization is a Monte Carlo simulated annealing procedure where the metastability plays the role of the energy and we use a simple temperature scheme.