Revealing Adverse Outcome Pathways from Public High-Throughput Screening Data to Evaluate New Toxicants by a Knowledge-Based Deep Neural Network Approach

Abstract

Traditional experimental testing to identify endocrine disruptors that enhance estrogenic signaling relies on expensive and labor-intensive experiments. We sought to design a knowledge-based deep neural network (k-DNN) approach to reveal and organize public high-throughput screening data for compounds with nuclear estrogen receptor α and β (ERα and ERβ) binding potentials. The target activity was rodent uterotrophic bioactivity driven by ERα/ERβ activations. After training, the resultant network successfully inferred critical relationships among ERα/ERβ target bioassays, shown as weights of 6521 edges between 1071 neurons. The resultant network uses an adverse outcome pathway (AOP) framework to mimic the signaling pathway initiated by ERα and identify compounds that mimic endogenous estrogens (i.e., estrogen mimetics). The k-DNN can predict estrogen mimetics by activating neurons representing several events in the ERα/ERβ signaling pathway. Therefore, this virtual pathway model, starting from a compound’s chemistry initiating ERα activation and ending with rodent uterotrophic bioactivity, can efficiently and accurately prioritize new estrogen mimetics (AUC = 0.864–0.927). This k-DNN method is a potential universal computational toxicology strategy to utilize public high-throughput screening data to characterize hazards and prioritize potentially toxic compounds.

Development, Validation, and Application of a Human Reproductive Toxicity Prediction Model Based on Adverse Outcome Pathway

“…In this study, a WoE-based consensus model (Figure S5) was applied to the data imputation process at tier 1 instead of selecting a model with the best performance. The consensus modeling step typically represents the best predictivity by leveraging the strengths of many ML algorithms. , Therefore, out of 88 models, 30 models with accuracy ≥ 0.70 and F1 measure ≥ 0.10 were used to build 11 consensus models in tier 1, and the performance of 11 consensus models was summarized in Table S8. Additionally, nine chemicals in the Hershberger assay have not only animal results, but also in chemico and in vitro results for one MIE and five KEs.…”

“…Therefore, the stacked Bayesian models for male and female reproductive toxicities greatly expanded the ADs by 1. 65 using an additional validation set collected from traditional reproductive toxicity guidelines. For male reproductive toxicity, a total of 17 male reproductive toxicants (Figure S13) were selected from the ToxRefDB database 6 via four principles (Materials and Methods).…”

Artificial Intelligence-Based Toxicity Prediction of Environmental Chemicals: Future Directions for Chemical Management Applications

“…In addition to using chemical structure, changes in biomolecular activity, cell activity, and toxicokinetics are also being used in this emerging field of AI-based toxicity prediction. In research on predicting changes in biomolecular activity, a large amount of toxicogenomics data and HTS data is used to associate with toxicity, which ultimately aims to predict the toxicity of chemicals by using an AI algorithm . Recent studies have mainly focused on hepatotoxicity as an end point based on the Open TG-GATEs database. , For example, Chen’s study developed a Tox-generative adversarial network (GAN) framework using transcriptomics data of Open TG-GATEs .…”

“…Baker et al 104 used a toxicity prediction model to identify the toxic mechanism of known cleft-palate-inducing chemicals and established a potential AOP. Finally, Ciallella et al 90 used a knowledge-based deep neural network approach to screen potential chemicals for virtual AOPs that link to rodent uterotrophic bioactivity via the ER signaling pathway.…”

Graph Attention Network Model with Defined Applicability Domains for Screening PBT Chemicals

“…Taking advantage of specific algorithms, QSAR models can first learn the correlation rules between the structural features (i.e., descriptors) and the chemical activities underlying the existing data (i.e., training data) and then utilize the learned rules efficiently to fill the data gaps . Deep learning (DL), an advancing field of artificial intelligence, witnessed prospects in QSAR modeling with large-scale database. , …”