Predicting Chemical Immunotoxicity through Data-Driven QSAR Modeling of Aryl Hydrocarbon Receptor Agonism and Related Toxicity Mechanisms

Abstract

Computational modeling has emerged as a time-saving and cost-effective alternative to traditional animal testing for assessing chemicals for their potential hazards. However, few computational modeling studies for immunotoxicity were reported, with few models available for predicting toxicants due to the lack of training data and the complex mechanisms of immunotoxicity. In this study, we employed a data-driven quantitative structure-activity relationship (QSAR) modeling workflow to extensively enlarge the limited training data by revealing multiple targets involved in immunotoxicity. To this end, a probe data set of 6,341 chemicals was obtained from a high-throughput screening (HTS) assay testing for the activation of the aryl hydrocarbon receptor (AhR) signaling pathway, a key event leading to immunotoxicity. Searching this probe data set against PubChem yielded 3,183 assays with testing results for varying proportions of these 6,341 compounds. 100 assays were selected to develop QSAR models based on their correlations to AhR agonism. Twelve individual QSAR models were built for each assay using combinations of four machine-learning algorithms and three molecular fingerprints. 5-fold cross-validation of the resulting models showed good predictivity (average CCR = 0.73). A total of 20 assays were further selected based on QSAR model performance, and their resulting QSAR models showed good predictivity of potential immunotoxicants from external chemicals. This study provides a computational modeling strategy that can utilize large public toxicity data sets for modeling immunotoxicity and other toxicity endpoints, which have limited training data and complicated toxicity mechanisms.

Figure 1

Computational modeling workflow of this study: (1) profiling the probe data set through PubChem, (2) bioassay selection, (3) QSAR modeling of selected bioassays, (4) model selection, (5) predictions of external chemicals.

Figure 2

Performance of the generated individual and consensus QSAR models, where models were trained with different combinations of ML algorithms and molecular fingerprints. (A) 5-fold cross-validation results for the AhR probe data set, evaluated using sensitivity, specificity, CCR, and PPV. (B) 5-fold cross-validation results for the 100 selected bioassays. Each point represents a QSAR model for an assay trained using the respective algorithm and fingerprint. The dashed line highlights the assay selection criteria of CCR > 0.7.

Figure 3

Predictions of 50 immunotoxicants by the consensus models of the 20 selected PubChem assays. PFOA - perfluorooctanoic acid; NDMA - N-nitrosodimethylamine; VCD - vinylcyclohexene dioxide; PFOS - perfluorooctanesulfonic acid; TCDD - 2,3,7,8-tetrachlorodibenzo-p-dioxin; p,p′-DDT - dichlorodiphenyltrichloroethane; 2,3,4,7,8-PCDF - 2,3,4,7,8-pentachlorodibenzofuran; 1,2,3,7,9-PCDF - 1,2,3,7,9-pentachlorodibenzofuran; 1,3,6,8-TCDF - 1,3,6,8-tetrachlorodibenzofuran.

Figure 4

Kernel density estimate (KDE) plot of the probability distribution of mean predictions from the Cosmetic Ingredients (CosIng) database and the Toxin and Toxin-Target Database (T3DB). The red dashed line represents the peak of the predictions of the immunotoxic chemical data set.