Predicting Chemical Ocular Toxicity Using a Combinatorial QSAR Approach

Abstract

Regulatory agencies require testing of chemicals and products to protect workers and consumers from potential eye injury hazards. Animal screening, such as the rabbit Draize test, for potential environmental toxicants is time-consuming and costly. Therefore, virtual screening using computational models to tag potential ocular toxicants is attractive to toxicologists and policy makers. We have developed quantitative structure–activity relationship (QSAR) models for a set of small molecules with animal ocular toxicity data compiled by the National Toxicology Program Interagency Center for the Evaluation of Alternative Toxicological Methods. The data set was initially curated by removing duplicates, mixtures, and inorganics. The remaining 75 compounds were used to develop QSAR models. We applied both k nearest neighbor and random forest statistical approaches in combination with Dragon and Molecular Operating Environment descriptors. Developed models were validated on an external set of 34 compounds collected from additional sources. The external correct classification rates (CCR) of all individual models were between 72 and 87%. Furthermore, the consensus model, based on the prediction average of individual models, showed additional improvement (CCR = 0.93). The validated models could be used to screen external chemical libraries and prioritize chemicals for in vivo screening as potential ocular toxicants.

Predictive Modeling of Estrogen Receptor Binding Agents Using Advanced Cheminformatics Tools and Massive Public Data

“…$$\begin{array}{c}\mathit{sensitivity}=\frac{\mathit{true}\mathit{positives}}{(\mathit{true}\mathit{positives}+\mathit{false}\mathit{negatives})}\\ \mathit{specificity}=\frac{\mathit{true}\mathit{negative}}{(\mathit{true}\mathit{negative}+\mathit{false}\mathit{positives})}\\ \mathit{CCR}=\frac{\mathit{sensitivity}+\mathit{specificity}}{2}\end{array}$$For the five-fold cross-validation procedures, the predictivity was similar across all the models (CCR = 0.642–0.749). However, the external predictions of the 264 unknown compounds showed a significant decrease in accuracy (CCR = 0.544–0.627), as observed in previous QSAR studies (Zhu et al, 2008a; Solimeo et al, 2012; Ng et al, 2015). Compared to individual models, the consensus model gave similar performance to the best individual models for both five-fold cross validation (sensitivity = 0.730, specificity = 0.704, and CCR = 0.717) and external predictions (sensitivity = 0.500, specificity = 0.683, and CCR = 0.592).…”

“…Each method was performed with both MOE and Dragon descriptors, as shown in the modeling workflow in Figure 1. The six resulting models (SVM-Dragon; SVM-MOE; RF-Dragon; RF-MOE; k NN-Dragon; and k NN-MOE) were averaged to give a consensus prediction, as described in previous publications (Solimeo et al, 2012; Kim et al, 2014). All models were validated using a five-fold cross validation.…”

Developing Enhanced Blood–Brain Barrier Permeability Models: Integrating External Bio-Assay Data in QSAR Modeling

“…However, the RF model has the best performance when predicting external compounds (R 2 =0.524 and MAE=0.399). The conflicts between the results obtained from cross-validation and external prediction were reported in many previous QSAR studies (21,36). Meanwhile, using AD did not show improvement of results for five-fold cross-validation or external set predictions.…”

“…For example, Digoxin (PubChem CID 30322), which is widely used in heart failure treatment, was proven to be actively transported out of the BBB by MDR1 (34). Excluding structural outliers from the modeling set may improve robustness of the QSAR models (35), while outliers in the external set should be detected by the model’s applicability domain (21,22). Since removing these structural outliers ( e.g .…”

“…An extra consensus QSAR model was then generated by averaging predictions of the three individual models. The development and application of consensus QSAR models have been reported in our previous publications (21–23). …”