CoMPARA: Collaborative Modeling Project for Androgen Receptor Activity.

Area of Science:

• Environmental Toxicology
• Computational Chemistry
• Endocrinology

Background:

• Endocrine disrupting chemicals (EDCs) are exogenous substances that interfere with hormonal systems.
• Concerns about adverse health effects necessitate robust methods for evaluating EDC bioactivity.
• High-throughput screening (HTS) and computational modeling are key approaches for EDC assessment.

Purpose of the Study:

• To virtually screen a large chemical dataset for potential androgen receptor (AR) activity.
• To develop and validate predictive models for AR binding, agonist, and antagonist activities.
• To support regulatory efforts like the U.S. Environmental Protection Agency's (EPA) Endocrine Disruptor Screening Program.

Main Methods:

• The Collaborative Modeling Project for Androgen Receptor Activity (CoMPARA) screened 55,450 chemical structures, building upon the Collaborative Estrogen Receptor Activity Prediction Project (CERAPP).
• Ninety-one predictive models were developed by 25 international groups using a training set of 1,746 chemicals from 11 HTS in vitro assays.
• Consensus models were created by combining individual model predictions to enhance accuracy.

Main Results:

• Consensus models achieved an average predictive accuracy of approximately 80% for AR activity when evaluated against curated literature data.
• The models were implemented in the OPERA application for screening new chemicals within a defined applicability domain.
• Predicted AR activities for EPA DSSTox database chemicals are publicly available.

Conclusions:

• Consensus modeling provides a reliable approach for predicting chemical endocrine activity, overcoming limitations of single models.
• The OPERA application and publicly accessible data facilitate the screening of chemicals for potential endocrine disruption.
• This work advances the assessment of EDCs, contributing to public and environmental health protection.