A method for in vitro data and structure curation to optimize for QSAR modelling of minimum absolute potency levels...

Area of Science:

• Toxicology
• Computational Chemistry
• Data Science

Background:

• Large-scale in vitro screening programs generate vast datasets relevant to human health.
• Quantitative Structure-Activity Relationship ((Q)SAR) modeling requires clearly defined endpoints and robust data.
• Existing datasets may contain variability affecting (Q)SAR model performance.

Purpose of the Study:

• To develop a comprehensive data curation procedure for interpreting in vitro experimental data.
• To enhance the reliability and accuracy of (Q)SAR models using high-quality screening data.
• To present a use case of the developed method for modeling Tox21 estrogen receptor α agonism assay data.

Main Methods:

• Implemented modules for selecting active compounds based on curve fitting, activity magnitude, and potency cut-offs.
• Included criteria for non-cytotoxicity at active concentrations and selection of highly pure substances.
• Developed a structure curation procedure for uniform representation of tautomeric classes and addressed assay signal interference.

Main Results:

• The data curation procedure systematically filters and refines in vitro screening datasets.
• Modeling Tox21 estrogen receptor α agonism data using the curated method showed improved performance compared to uncurated data.
• The developed procedure enhances the quality of data used for (Q)SAR development.

Conclusions:

• A robust data curation methodology is essential for reliable (Q)SAR development from large-scale screening programs.
• The presented procedure effectively improves the quality of in vitro data for predictive toxicology.
• This approach supports the accurate assessment of chemical safety and human health risks.