Visualizing and identifying conformational ensembles in molecular dynamics trajectories

Simulating the dynamics of complex biomolecules produces trajectories comprising a large number of different configurations of the molecule. These configurations must be classified into a small number of conformational ensembles representing essential changes in the shape of the molecule. Using a conformational distance measure based on the changes in intramolecular atom distances, we show that these trajectories can be visualized efficiently by a planar map such that different conformational ensembles appear as well-separated point sets. We then use statistical cluster analysis to identify clusters that represent different conformational ensembles.