Search

Modeling approaches such as quantitative structure-activity relationships (QSARs) use molecular descriptors to predict the bioavailable properties of a compound in biota. However, these models have mainly been derived based on empirical data for lipophilic neutral compounds and may not predict the uptake of ionizable compounds. The majority of pharmaceuticals are ionizable, and freshwaters can have a range of pH values that affect speciation. In this study, we assessed the uptake of 10 pharmaceuticals (acetazolamide, beclomethasone, carbamazepine, diclofenac, gemfibrozil, ibuprofen, ketoprofen, norethindrone, propranolol, and warfarin) with differing modes of action and physicochemical properties (p K a , log S, log D, log K ow , molecular weight (MW), and polar surface area (PSA)) by an in vitro primary fish gill cell culture system (FIGCS) for 24 h in artificial freshwater. Principal component analysis (PCA) and partial least-squares (PLS) regression was used to determine the molecular descriptors that influence the uptake rates. Ionizable drugs were taken up by FIGCS; a strong positive correlation was observed between log S and the uptake rate, and a negative correlation was observed between p K a , log D, and MW and the uptake rate. This approach shows that models can be derived on the basis of the physicochemical properties of pharmaceuticals and the use of an in vitro gill system to predict the uptake of other compounds. There is a need for a robust and validated model for gill uptake that could be used in a tiered risk assessment to prioritize compounds for experimental testing.

Modeling and prediction of polar organic chemical integrative sampler (POCIS) sampling rates (Rs) for 73 compounds using artificial neural networks (ANNs) is presented for the first time. Two models were constructed: the first was developed ab initio using a genetic algorithm (GSD-model) to shortlist 24 descriptors covering constitutional, topological, geometrical and physicochemical properties and the second model was adapted for Rs prediction from a previous chromatographic retention model (RTD-model). Mechanistic evaluation of descriptors showed that models did not require comprehensive a priori information to predict Rs. Average predicted errors for the verification and blind test sets were 0.03 ± 0.02 L d(-1) (RTD-model) and 0.03 ± 0.03 L d(-1) (GSD-model) relative to experimentally determined Rs. Prediction variability in replicated models was the same or less than for measured Rs. Networks were externally validated using a measured Rs data set of six benzodiazepines. The RTD-model performed best in comparison to the GSD-model for these compounds (average absolute errors of 0.0145 ± 0.008 L d(-1) and 0.0437 ± 0.02 L d(-1), respectively). Improvements to generalizability of modeling approaches will be reliant on the need for standardized guidelines for Rs measurement. The use of in silico tools for Rs determination represents a more economical approach than laboratory calibrations.

The computational generation of gradient retention time data for retrospective detection of suspected sports doping species in postanalysis human urine sample data is presented herein. Retention data for a selection of 86 compounds included in the London 2012 Olympic and Paralympic Games drug testing schedule were used to train, verify, and test a range of computational models for this purpose. Spiked urine samples were analyzed using solid phase extraction followed by ultrahigh-pressure gradient liquid chromatography coupled to electrospray ionization high-resolution mass spectrometry. Most analyte retention times varied ≤0.2 min over the relatively short runtime of 10 min. Predicted retention times were within 0.5 min of experimental values for 12 out of 15 blind test compounds (largest error: 0.97 min). Minimizing the variance in predictive ability across replicate networks of identical architecture is presented for the first time along with a quantitative discussion of the contribution of each selected molecular descriptor toward the overall predicted value. The performance of neural computing predictions for isobaric compound retention time is also discussed. This work presents the application of neural networks to the prediction of gradient retention time in archived high-resolution urine analysis sample data for the first time in the field of anti-doping.