The "nearest single neighbor" method-finding families of conformations within a sample

A simple method for self-organization of conformation samples into families is presented. According to this method, any large sample of molecular conformations may be reorganized according to the nearest single root-mean-square displacement (rmsd) neighbor, starting at any chosen "seed" conformation. Following such reordering, conformational families may be determined by a novel process that maximizes family sizes while minimizing family mixing. This process eliminates much of the arbitrariness that was inherent in most of the related methods of conformation clustering. We demonstrate the construction of rmsd matrices and discuss the convergence criteria for the sample size as well as criteria for determining the cutoff value for the definition of families in each sample. The method is invariant to changes of the "seed" conformation. After applying this method, families of conformations may be more easily recognized in graphic matrices. The method has been applied to the analysis of the conformational space of two cyclic peptides. It is also shown that the "organized" conformational space, at least in those specific examples, has an energy topology that reminds of energy basins. The method is general and applicable to molecules of any type.