MULTICONFORMATIONAL INVESTIGATIONS OF POLYPEPTIDIC STRUCTURES, USING CLUSTERING METHODS AND PRINCIPAL COMPONENTS-ANALYSIS

Recently, we have applied random-search and energy minimization proced ures in the framework of molecular mechanics in order to investigate t he structural properties of polypeptides. Consequently, populations of conformers were generated and analyzed. The problem of the identifica tion of conformation domains was addressed using factorial analysis su ch as clustering methods and principal components analysis. A strategy to apply such techniques to conformational populations is presented h ere. The results obtained for proline-containing peptides demonstrate the existence of five families of conformations, describing completely the generated population. The homogeneity and variability within fami lies are discussed in terms of their characteristic molecular structur es.