The present work employs 152 benzene-carboxylic acid' esters having computed the toxicity within the range [2.251, 10.222] for fathead minnow fish (Pimephales promelas). Calibration set includes many pairs having very similar chemical structure, size, shape and hydrophilicity, but very different value of ECOSAR toxicity or vice versa. The QSTR study, which uses all esters as calibration set, emphasized a large percent (16.2%) of outliers. In this QSTR study most of the estimated values of toxicity for outliers are much lower than ECOSAR toxicity. The LogP and some aromaticity descriptors are predictors. The best QSTR for esters having low value (< 5.5) of ECOSAR toxicity and the best QSTR for esters having high value (> 5.5) of ECOSAR toxicity are obtained when the number of outliers is very small. These QSTRs are different enough and highlight opposite influences of certain descriptors on toxicity. The results emphasize two possibilities: (a) the esters having low value of ECOSAR toxicity and the esters having high value of ECOSAR toxicity are included in two different classes from the point of view of structure-toxicity relationship and/or (b) many high values of ECOSAR toxicity are wrong. By comparison, a QSTR using experimental values of toxicity against rats for 37 benzene-carboxylic esters included in the same database gives good correlation experimental/computed values of toxicity, the number of outliers is null and the result of validation test is good.
Keywords: Benzene-carboxylic acid esters, ECOSAR toxicity, fathead minnow, QSTR, PRECLAV.