The external prediction capability of quantitative structure-activity relationship (QSAR) models is often quantified using the predictive squared correlation coefficient, q (2). This index relates the predictive residual sum of squares, PRESS, to the activity sum of squares, SS, without postprocessing of the model output, the latter of which is automatically done when calculating the conventional squared correlation coefficient, r (2). According to the current OECD guidelines, q (2) for external validation should be calculated with SS referring to the training set activity mean. Our present findings including a mathematical proof demonstrate that this approach yields a systematic overestimation of the prediction capability that is triggered by the difference between the training and test set activity means. Example calculations with three regression models and data sets taken from literature show further that for external test sets, q (2) based on the training set activity mean may become even larger than r (2). As a consequence, we suggest to always use the test set activity mean when quantifying the external prediction capability through q (2) and to revise the respective OECD guidance document accordingly. The discussion includes a comparison between r (2) and q (2) value ranges and the q (2) statistics for cross-validation.

Quantitative and qualitative structure-activity relationships (QSARs) have a great potential to support the risk assessment of chemicals, provided there are tools available that allow evaluation of the suitability of QSARs for the compounds of interest. In this context, a pragmatic approach is to discriminate excess toxicity from narcotic effect levels, because the latter can be estimated from QSARs and thus have a low priority for experimental testing. To develop a respective scheme for the acute daphnid toxicity as one of the primary ecotoxicological endpoints, 1067 acute toxicity data entries for 380 chemicals involving the daphnid species Daphnia magna were taken from the on-line literature, and quality checks such as water solubility were employed to eliminate apparently odd data entries. For 36 known narcotics with LC50 values referring to D. magna, a reference baseline QSAR is derived. Compounds with LC50 values above a certain threshold defined relative to their predicted baseline toxicity are classified as exerting excess toxicity. Three simple discrimination schemes are presented that enable the identification of excess toxicity from structural alerts based on the presence or absence of certain heteroatoms and their chemical functionality. Moreover, a two-step classification approach is introduced that enables a prioritization of organic compounds with respect to their need for experimental testing. The discussion includes reaction mechanisms that may explain the association of structural alerts with excess toxicity, a comparison with predictions derived from mode of action-based classification schemes, and a statistical analysis of the discrimination performance in terms of detailed contingency table statistics.

No abstract

No abstract