ALGORITHMS FOR CLUSTERING MOLECULAR-DYNAMICS CONFIGURATIONS

Two traditional clustering algorithms are applied to configurations fr om a long molecular dynamics trajectory and compared using two sets of test data. First, a subset of atoms was chosen to present conformatio ns which naturally fall into a number of clusters. Second, a subset of atoms was selected to span a relatively continuous region of conforma tional space rather than form discrete conformational classes. Of the two algorithms used, the single linkage method is inappropriate for th is kind of data. The divisive hierarchical method, based on minimizing the difference between cluster centroids and extrema, is successful b ut also prone to imposing clustering hierarchy where none can be justi fied. (C) 1994 by John Wiley and Sons, Inc.