High-Throughput Screening to Predict Chemical-Assay Interference.

Area of Science:

• Toxicology
• Computational Chemistry
• Biochemistry

Background:

• High-throughput screening (HTS) assays are crucial for toxicology but susceptible to chemical interference, leading to false positives.
• Luciferase and fluorescence-based readouts commonly used in HTS can be affected by specific chemical structures.
• The Toxicology in the 21st Century (Tox21) program generates extensive HTS data, necessitating methods to ensure data reliability.

Purpose of the Study:

• To identify and characterize chemical interference in Tox21 HTS assays.
• To develop predictive models for assay interference using chemical structures and properties.
• To create a tool (InterPred) to help researchers predict and mitigate false positives in HTS data.

Main Methods:

• Assays were conducted to measure luciferase inhibition and autofluorescence across various conditions.
• Chemical structures were analyzed using self-organizing maps and hierarchical clustering.
• Machine learning algorithms were employed to build predictive models based on molecular descriptors.

Main Results:

• Out of 8,305 chemicals tested, interference rates ranged from 0.5% (red autofluorescence) to 9.9% (luciferase inhibition).
• Predictive models achieved accuracies of approximately 80% in forecasting assay interference.
• Structural clusters were successfully correlated with specific interference activity profiles.

Conclusions:

• Chemical interference is a significant factor in HTS toxicology assays.
• Machine learning models can effectively predict assay interference, improving data quality.
• The InterPred tool aids in reducing false positives, thereby increasing confidence in HTS results.