Development of a Predictive Model for N-Dealkylation of Amine Contaminants Based on Machine Learning Methods

Area of Science:

• Environmental Chemistry
• Computational Toxicology
• Biotechnology

Background:

• Amines are prevalent environmental pollutants with potential health risks.
• Cytochrome P450-mediated N-dealkylation influences amine metabolism and safety.
• Current methods for screening amine N-dealkylation are inefficient for high-throughput analysis.

Purpose of the Study:

• To develop machine learning models for predicting the N-dealkylation of amine pollutants.
• To establish a high-throughput screening method for identifying critical biotransformation pathways of emerging amine contaminants.
• To assess the performance of various machine learning algorithms in classifying N-dealkylation potential.

Main Methods:

• Compiled a dataset of 286 emerging amine pollutants from literature and databases.
• Applied four machine learning algorithms: random forest, gradient boosting decision tree, extreme gradient boosting, and multi-layer perceptron.
• Utilized seven molecular descriptors representing reactivity-fit and structural-fit for model development.
• Developed an ensemble model integrating three algorithms using a consensus strategy.

Main Results:

• Extreme gradient boosting achieved the highest prediction accuracy of 81.0%.
• The SlogP_VSA2 descriptor was identified as the most influential factor for predicting N-dealkylation.
• The ensemble model demonstrated superior performance with an accuracy rate of 86.2%.
• The developed models provide a reliable method for classifying N-dealkylation potential.

Conclusions:

• Machine learning offers a powerful tool for the high-throughput screening of amine N-dealkylation.
• The developed classification models can significantly aid in assessing the metabolic safety of amine pollutants.
• This study provides methodological support for environmental risk assessment of emerging amine contaminants.