ABSTRACT
Simple and multiple linear regressions were applied to the development of fish toxicity QSAR models for the 96-h LC50 to the fathead minnow, Pimephales promelas. The data on unbranched saturated primary alkylamines as well as the complete data set were well-fitted to linear QSAR models using log P or the valence first-order connectivity index (1XV) as descriptors. Although adding data on other subclasses of amines in this data set yield acceptable QSARs, only the tertiary amine subclass provided a poor fit with both of these descriptors. The amines include both acyclic and cyclic derivatives, either with no additional functional groups, or with the hydroxyl, keto, methoxy, and propargyl moieties. The molecular mechanism for fish toxicity of these amines as well as the outliers in the study were investigated. Based upon the calculated log P value of -1.40, tripropargylamine has an apparent excess toxicity of 84 times; in contrast, the measured shake-flask log P for this compound was subsequently found to be 1.26, giving a predicted LC50 consistent with the observed value. An upward curvature of the QSAR plot for the most hydrophilic compounds suggests a shift in mechanism for the lowest members of the series.