Principal coordinate analysis on a protein model

A well-studied 46-bead protein model is the vehicle for examining principal coordinate analysis as a tool for interpreting topographies of complex pot ential surfaces. This study compares the effectiveness of several definitio ns of the comparison variable for revealing information about topographies. The extent of the information is ascertained by comparing the results of t he various forms of principal coordinate analysis with results obtained fro m construction of interconnected monotonic sequences of linked stationary p oints (IMSLiSP) on the same surface. The conclusion is that the most powerf ul formulation of principal coordinate analyses for understanding protein f olding and, in general, topographies of complex potentials, uses the change s in the set of interparticle distances as the definition of the comparison vector. However, even with this choice, the more efficient principal coord inate analysis is not able to reveal the extent of information contained in a more cumbersome IMSLiSP analysis. (C) 1999 American Institute of Physics . [S0021-9606(99)51419-0].