No abstract

An evaluated database of whole body in vivo biotransformation rate estimates in fish was used to develop a model for predicting the primary biotransformation half-lives of organic chemicals. The estimated biotransformation rates were converted to half-lives and divided into a model development set (n=421) and an external validation set (n=211) to test the model. The model uses molecular substructures similar to those of other biodegradation models. The biotransformation half-life predictions were calculated based on multiple linear regressions of development set data against counts of 57 molecular substructures, the octanol-water partition coefficient, and molar mass. The coefficient of determination (r2) for the development set was 0.82, the cross-validation (leave-one-out coefficient of determination, q2) was 0.75, and the mean absolute error (MAE) was 0.38 log units (factor of 2.4). Results for the external validation of the model using an independent test set were r2 = 0.73 and MAE = 0.45 log units (factor of 2.8). For the development set, 68 and 95% of the predicted values were within a factor of 3 and a factor of 10 of the expected values, respectively. For the test (or validation) set, 63 and 90% of the predicted values were within a factor of 3 and a factor of 10 of the expected values, respectively. Reasons for discrepancies between model predictions and expected values are discussed and recommendations are made for improving the model. This model can predict biotransformation rate constants from chemical structure for screening level bioaccumulation hazard assessments, exposure and risk assessments, comparisons with other in vivo and in vitro estimates, and as a contribution to testing strategies that reduce animal usage.

Abstract-A compound's bioconcentration factor (BCF) is the most commonly used indicator of its tendency to accumulate in aquatic organisms from the surrounding medium. Because it is expensive to measure, the BCF is generally estimated from the octanol/water partition coefficient (K ow ), but currently used regression equations were developed from small data sets that do not adequately represent the wide range of chemical substances now subject to review. To develop an improved method, we collected BCF data in a file that contained information on measured BCFs and other key experimental details for 694 chemicals. Log BCF was then regressed against log K ow and chemicals with significant deviations from the line of best fit were analyzed by chemical structure. The resulting algorithm classifies a substance as either nonionic or ionic, the latter group including carboxylic acids, sulfonic acids and their salts, and quaternary N compounds. Log BCF for nonionics is estimated from log K ow and a series of correction factors if applicable; different equations apply for log K ow 1.0 to 7.0 and Ͼ7.0. For ionics, chemicals are categorized by log K ow and a log BCF in the range 0.5 to 1.75 is assigned. Organometallics, nonionics with long alkyl chains, and aromatic azo compounds receive special treatment. The correlation coefficient (r 2 ϭ 0.73) and mean error (0.48) for log BCF (n ϭ 694) indicate that the new method is a significantly better fit to existing data than other methods.

No abstract

No abstract