Machine learning for mixture toxicity analysis based on high-throughput printing technology.

Area of Science:

• Environmental Toxicology
• Computational Toxicology
• Biotechnology

Background:

• Conventional mixture toxicity (Mix-tox) testing is resource-intensive, limiting the scope of predictive models.
• Machine learning (ML) offers potential for accelerating toxicity analysis but requires substantial toxicological data.
• A significant challenge in ML-driven toxicology is the scarcity of comprehensive, high-quality datasets.

Purpose of the Study:

• To develop a standardized high-throughput experimental strategy for comprehensive Mix-tox analysis.
• To address the challenge of limited toxicology big-data for ML model development.
• To build and train an ML model for accurate prediction of mixture toxicity effects.

Main Methods:

• Designed a big-sample dataset strategy for high-throughput Mix-tox analysis.
• Developed and trained a random forest algorithm using concentration variates as input.
• Utilized bio-luminescent inhibition rate as the output metric for toxicity assessment.

Main Results:

• Successfully implemented a full strategy for Mix-tox analysis, from data generation to prediction.
• A well-trained random forest model demonstrated high accuracy in assessing mixture toxicity.
• The approach effectively utilized concentration data to predict bio-luminescent inhibition rates.

Conclusions:

• The developed high-throughput experimental strategy facilitates robust Mix-tox analysis.
• The random forest model proves valuable for predicting mixture toxicity, overcoming data limitations.
• This approach supports the broader adoption of ML in toxicology for efficient chemical safety assessment.