Prediction of Luminescent Bacteria Toxicity Using Quantum Chemical Descriptors:Test of a Classification Scheme

摘要

The classification scheme of Verhaar et al. (1992) based on fish toxicity data wasinvestigated as to whether it is applicable to photobacteria toxicity data (Microtox). For thispurpose 240 diverse chemicals were classified according to this scheme into 3 classes: 1) nonpo-lar narcotics, 2) polar narcotics, and 3) reactive chemicals. In good agreement with former re-sults, a baseline quantitative structure-activity relationship (QSAR) for the photobacteria tox-icity was established: log 1/EC50 = 1.07~*logP—2.36. Results from the analysis of relativetoxicity showed that the classification scheme is not applicable to the Microtox data withoutmodification. Many of the Class 1 compounds are considerably more toxic than the baselinemodel predicts. Despite this, a rather good QSAR with molecular polarizability, lowest unoccu-pied molecular orbital (LUMO) energy, and a shape descriptor was found for the Class 1 com-pounds. For Classes 2 and 3, the relative toxicities indicate a mixing of polar narcotic andreactive compounds. Also the corresponding QSARs were rather poor. For Class 2, however, theresult could be improved substantially by excluding some outliers. Maybe some minor modifi-cations of the rules are sufficient to adapt them to the Microtox data. For Class 3, further classi-fication into reactive subclasses is required. The predictability of the Class 1 QSAR for the other2 classes is relatively high, revealing that many Class 2 and 3 compounds are similar to nonpo-lar narcotics.