Steven Hayward
A straightforward method for the determination of dynamic domains from pairs of conformations or general conformational ensembles of proteins is presented. The method applies metric multi-dimensional scaling (MDS) to distance-differences or root-mean square-fluctuations of inter-atomic distances (RMSFIDs). The approach determines points in a low-dimensional space, each representing an amino-acid residue, where distances between the points give an approximation to the distance-differences, in the case of a pair of conformations, or RMSFIDs for an ensemble. This point-based representation enables top-down clustering methods to be used to determine dynamic domains. The two implementations, Pair-DD and Ensemble-DD, are demonstrated on idealized examples where domains move as perfect rigid bodies, on conformational pairs and ensembles derived from X-ray structures both monomeric and multimeric, and on trajectories derived from simulation methods. A parameter is proposed which can be used as a threshold for acceptance of dynamic domains to enable automatic assignment. The results show excellent correspondence with a well-established approach, but the method has the added advantage of being versatile in that it is applicable to both a pair of structures and an ensemble of conformations. Furthermore, for a pair of conformations, a one-dimensional MDS coordinate seems to be sufficient, meaning that the degree of association of a residue with a dynamic domain can be visualized in a simple plot.