Development of human biotransformation QSARs and application for PBT assessment refinement

Evaluating consumer exposure to disinfecting chemicals against coronavirus disease 2019 (COVID-19) and associated health risks

“…Here, we should confess that QSAR prediction of biotransformation half-lives of quaternary ammonium salts is challenging and uncertain because they are seldom included in the training sets for developing the currently used QSAR tools. While the QSAR used in this work ( Papa et al, 2018 ) recognizes that C12BAC and C14BAC fall within its applicability domain and warns that C16BAC and C18BAC exceed it, readers should be with caution when interpreting the predicted blood concentrations of quaternary ammonium salts. Nevertheless, our predicted relative order of elimination rates of the investigated disinfecting chemicals seems to agree with experimental studies and thus be somewhat reasonable.…”

“…The model uses default values of internal energies and activation energies to adjust the partition coefficients and reaction rate constant to a certain temperature. In addition, human biotransformation-halves are computed using the human biotransformation QSAR models by Papa et al (2018) implemented in the QSARINS-Chem module ( Gramatica et al, 2014 ) of the software QSAR-INSubria (QSARINS) ( Gramatica et al, 2013 ). Since the training set in Papa et al models contains few ionogenic organic chemicals, we calculate both the best estimates and 95th percentiles to characterize the uncertainty associated with the predicted biotransformation halves.…”

QSARINS‐Chem standalone version: A new platform‐independent software to profile chemicals for physico‐chemical properties, fate, and toxicity

“…These models are useful to fill data gaps, for virtual screenings, and/or for the identification of safer alternatives to unsafe pollutants. Furthermore, the availability of multiple models, which can be combined to generate consensus predictions, helps to reduce the uncertainty associated with the prediction of a single property/activity, they can cross‐validate in silico predictions and experiments, and support decision making processes 4–13 …”

Intercorrelation Limits in Molecular Descriptor Preselection for QSAR/QSPR

“…The entire process is called variable reduction according to terms of Todeschini and Consonni: “Variable reduction consists of the selection of a subset of variables able to preserve the essential information contained in the whole dataset but eliminating redundancy, too highly correlated variables, etc. Variable reduction differs from variable selection in the fact that the subset of variables is selected independently from the response of interest.” While the removal of constant (or nearly constant) variables is relatively straightforward, there is no clear consensus on the choice of the specific intercorrelation limit in QSAR studies: a random sample from the QSAR literature of the recent years reveals choices of 1.000, 0.98, 0.95, 0.90, 0.80, and even 0.70 . Moreover, most studies either do not report the selected intercorrelation limit, or simply omit this step.…”