NEW PRINCIPAL COMPONENTS DERIVED PARAMETERS DESCRIBING MOLECULAR DIVERSITY OF HETEROAROMATIC RESIDUES

This study reports on determination of new descriptors for heteroaroma tic ring fragments permitting to take into account differences in anch oring positions of such substructures in a given molecule. For a set o f 72 aromatic moieties (five- and six-membered monocyclic and benzo fu sed heteroaromatics containing one or two heteroatoms) parameters incl uding 121 2D physico-chemical descriptors, six receptor surface models descriptors, and three 3D variables have been calculated and new sets of QSAR parameters were generated from principal component analyses o f the data matrix this obtained. Hierarchical clustering of the obtain ed principal components as well as applying both D-optimal design and a new function for molecular estimation allowed partitioning of the he teroaromatic substituents into similar families and permitted a classi fication of the structures according to their molecular similarity. Th e descriptors obtained finally were tested for their prediction capaci ty using a previously reported dataset of biologically active compound s. The results obtained hereby clearly demonstrate the usefulness of t he new principal properties.