ADMETopt2

ADMETopt2 optimizes chemical structures to reduce Ames mutagenicity by deriving and assessing structural transformation rules using matched molecular pairs analysis and machine learning.

Key Features:

• Integration of Large Datasets: Leverages an Ames mutagenicity dataset of 8,576 compounds to derive transformation rules aimed at reversing mutagenicity.
• Matched Molecular Pairs Analysis: Applies matched molecular pairs (MMP) analysis to identify and evaluate structural changes that influence mutagenic properties.
• Consensus Machine Learning Model: Uses a consensus model to assess generalizability and validity of transformation rules, reporting 0.815 accuracy on external validation sets and a reasonable applicability domain.
• Transformation Rules Assessment: Provides detailed evaluation of mutagenicity transformation rules and highlights the importance of the local chemical environment at attachment points for successful transformations.

Scientific Applications:

• Drug Discovery Optimization: Guides structural modification of compounds during early-stage drug development to reduce Ames mutagenicity.
• Toxicity Endpoint Extension: Enables extension of the derived methodologies and transformation rules to other toxicity endpoints.
• Empirical Rule Enhancement: Enhances traditional medicinal chemistry empirical rules with data-driven transformation rules from large-scale analysis.

Methodology:

Uses an 8,576-compound Ames mutagenicity dataset, matched molecular pairs (MMP) analysis to extract transformation rules, and a consensus machine learning model validated on external sets (accuracy 0.815) to assess and generalize the rules.