NR-ToxPred

NR-ToxPred predicts binding interactions of small molecules with human nuclear receptors to identify potential endocrine-disrupting chemicals and related activity types.

Key Features:

• Machine Learning-Based Predictions: An ensemble of machine learning models produces classification predictions for molecular interactions with nuclear receptors.
• Training Dataset: Models are trained on the Nuclear Receptor Activity (NuRA) dataset comprising experimentally derived small-molecule activities.
• Predicted Interaction Types: The tool predicts agonism, antagonism, binding, and effector binding for chemical–receptor pairs.
• Target Scope: Predictions cover nine different human nuclear receptors.
• Applicability Domain: An applicability domain is defined using Tanimoto similarity to molecules in the training set to constrain reliable prediction space.

Scientific Applications:

• Toxicology Screening: Rapid in silico screening and prioritization of chemicals for downstream biological testing in toxicology studies.
• Endocrine Disruptor Assessment: Identification and prioritization of potential endocrine-disrupting chemicals (EDCs) based on predicted interactions with nuclear receptors.
• Pharmacology and Drug Discovery: Prioritization of small molecules for drug discovery efforts targeting nuclear receptors by predicting modes of action such as agonism or antagonism.
• Regulatory Prioritization: Support for regulatory assessment workflows by enabling high-throughput computational prioritization of compounds for further evaluation.

Methodology:

An ensemble of machine learning classifiers was trained on the Nuclear Receptor Activity (NuRA) dataset to predict agonism, antagonism, binding, and effector binding across nine human nuclear receptors, with an applicability domain defined by Tanimoto similarity to training molecules.