TIRESIA: An eXplainable Artificial Intelligence Platform for Predicting Developmental Toxicity

Abstract

Herein, a robust and reproducible eXplainable Artificial Intelligence (XAI) approach is presented, which allows prediction of developmental toxicity, a challenging human-health endpoint in toxicology. The application of XAI as an alternative method is of the utmost importance with developmental toxicity being one of the most animal-intensive areas of regulatory toxicology. In this work, the established CAESAR (Computer Assisted Evaluation of industrial chemical Substances According to Regulations) training set made of 234 chemicals for model learning is employed. Two test sets, including as a whole 585 chemicals, were instead used for validation and generalization purposes. The proposed framework favorably compares with the state-of-the-art approaches in terms of accuracy, sensitivity, and specificity, thus resulting in a reliable support system for developmental toxicity ensuring informativeness, uncertainty estimation, generalization, and transparency. Based on the eXtreme Gradient Boosting (XGB) algorithm, our predictive model provides easy interpretative keys based on specific molecular descriptors and structural alerts enabling one to distinguish toxic and nontoxic chemicals. Inspired by the Organisation for Economic Co-operation and Development (OECD) principles for the validation of Quantitative Structure–Activity Relationships (QSARs) for regulatory purposes, the results are summarized in a standard report in portable document format, enclosing also details concerned with a density-based model applicability domain and SHAP (SHapley Additive exPlanations) explainability, the latter particularly useful to better understand the effective roles played by molecular features. Notably, our model has been implemented in TIRESIA (Toxicology Intelligence and Regulatory Evaluations for Scientific and Industry Applications), a free of charge web platform available at .

Compound Classification and Consideration of Correlation with Chemical Descriptors from Articles on Antioxidant Capacity Using Natural Language Processing

“…Additionally we found dexcriptions of highly toxic substances, known as pro-oxidant or pro-inflammatory substances . Other studies have also supported this finding. , It is known that highly toxic substances tend to have a low “qed” value. , Therefore, it can be said that the values of “qed” are clearly different from those of high “qed” substances used as pharmaceuticals, and the distribution of “qed” is considered to be more extensive than this. Furthermore, the mean values for certain descriptors varied significantly across clusters.…”

“…41,42 It is known that highly toxic substances tend to have a low "qed" value. 24,43 Therefore, it can be said that the values of "qed" are clearly different from those of high "qed" substances used as pharmaceuticals, and the distribution of "qed" is considered to be more extensive than this. Furthermore, the mean values for certain descriptors varied significantly across clusters.…”

CIRCE: Web-Based Platform for the Prediction of Cannabinoid Receptor Ligands Using Explainable Machine Learning

“…Specifically, four cooperating classification models were generated for predicting and rationalizing molecular determinants driving selective ligand binding to CB 1 R and CB 2 R. Each classification model was independently derived based on different sets of training data carefully curated from the ChEMBL database (release 31) . These compound pools contain a wealth of chemical information along with high-quality experimental data for binding to CB 1 R and CB 2 R. A new substructure-based core-substituent fingerprint (CSFP) was used to encode the structural information, and SHAP values − were computed to explain individual predictions. Eventually, the four models were assembled to build a multilayer classifier, which is freely accessible through a web platform designated Cannabinoid Iterative Revaluation for Classification and Explanation (CIRCE) that is provided with a user-friendly graphical interface. , CIRCE returns predictions on demand and instantly provides a detailed portable report of prediction outcomes.…”

Interpretation of Drop Size Predictions from a Random Forest Model Using Local Interpretable Model-Agnostic Explanations (LIME) in a Rotating Disc Contactor

“…Younisse et al 8 provided Shap explanations for anomaly detection. Togo et al 9 provided an Explainable framework for toxicity prediction. Jang et al 10 5 and has become very popular in prediction of complex machine learning models.…”

“…provided Shap explanations for anomaly detection. Togo et al provided an Explainable framework for toxicity prediction. Jang et al augmented Fault Diagnosis of Industrial Processes modeling with SHAP explanations.…”