NOVEL GRAPH THEORETICAL APPROACH TO HETEROATOMS IN QUANTITATIVE STRUCTURE - ACTIVITY RELATIONSHIPS

A novel approach to characterization of heteroatoms in graph theoretical approaches to quantitative structure-activity relationships (QSAR) is outlined. The basis of the approach is the use of diagonal entries of the adjacency matrix as variable parameter, in full analogy to the well known generalization of the Huckel Molecular Orbitals (HMO) method when extended to heterconjugated systems. The approach is illustrated on clonidine-like compounds where carbon and chlorine atoms are discriminated by using x = -0.20 as the diagonal entry for chlorine atoms. Derived weighted path numbers are used as descriptors and a multiple regression based on three descriptors resulted in the correlation coefficient R = 0.977 and the standard error S = 0.233. This represents a substantial improvement over the best traditional QSAR analysis which involves five descriptors (in a nonlinear correlation equation with R = 0.964 and S = 0.301). A detailed comparison is made with available QSAR results, and the advantages (as well as limitations) of graph theoretical descriptors are discussed.