High-throughput screening of chemicals as functional substitutes using structure-based classification models.

Area of Science:

• Computational chemistry
• Green chemistry
• Toxicology

Background:

• Green chemistry aims to identify functional chemicals with low human and environmental impact.
• Formulation chemists and engineers face challenges in finding suitable chemical substitutes.
• Existing methods for identifying functional chemical substitutes are often limited.

Purpose of the Study:

• To develop and apply a machine learning methodology for identifying potential chemical functional substitutes.
• To create harmonized function categories for modeling chemical uses.
• To screen large chemical libraries for safer alternatives using quantitative structure-use relationship (QSUR) models.

Main Methods:

• Collected and analyzed publicly available data on chemical functions in products and processes.
• Developed 41 quantitative structure-use relationship (QSUR) models using random forest classification based on chemical descriptors.
• Screened a library of ~6400 chemicals for functional substitutes and merged results with high-throughput (HT) bioactivity screening data.

Main Results:

• Identified uses for 3121 chemicals, with 4412 predicted functional uses having >80% probability.
• Successfully applied QSUR models to screen Tox21 chemicals, identifying over 1600 candidate chemical alternatives.
• Demonstrated a high-throughput (HT) screening approach for greener chemical alternatives by combining functional use and hazard data.

Conclusions:

• The developed QSUR models provide an efficient method for identifying functional chemical substitutes.
• This approach facilitates high-throughput (HT) screening for greener chemical alternatives by integrating functional use and toxicity data.
• The methodology can be rapidly applied to large chemical datasets to accelerate the discovery of safer chemicals.