Carcinogenic classification of polycyclic aromatic hydrocarbons through theoretical descriptors

摘要

Polycyclic aromatic hydrocarbons (PAHs) constitute an important family of molecules capable of inducing chemical carcinogenesis. In this work we report a comparative structure-activity relationship (SAR) study for 81 PAHs using different methodologies. The recently developed electronic indices methodology (EIM) with quantum descriptors obtained from different semiempirical methods (AM1, PM3, and PM5) was contrasted against more standard pattern recognition methods (PRMs), principal component analysis (PCA), hierarchical cluster analysis (HCA), Kth nearest neighbor (KNN), soft independent modeling of class analogies (SIMCA), and neural networks (NN). Our results show that PRMs validate the statistical value of electronic parameters derived from EIM analysis and their ability to identify active compounds. EIM outperformed more standard SAR methodologies and does not appear to be significantly Hamiltonian-dependent. (c) 2005 Wiley Periodicals, Inc.