The ToxCast pipeline: updates to curve-fitting approaches and database structure

Area of Science:

• Toxicology and Cheminformatics
• Computational Biology and Bioinformatics
• Environmental Health Sciences

Background:

• The US Environmental Protection Agency Toxicity Forecaster (ToxCast) program generates extensive in vitro screening data for chemical hazard assessment.
• The ToxCast data pipeline (tcpl) R package and invitrodb MySQL database manage and analyze this high-throughput screening data.
• Existing curve-fitting models in tcpl were limited for comprehensive toxicological profiling.

Purpose of the Study:

• To describe major updates to the tcpl R package and invitrodb database.
• To incorporate a broader range of curve-fitting models for enhanced analysis of ToxCast data.
• To improve the consistency, reproducibility, and utility of in vitro toxicology data.

Main Methods:

• Expanded tcpl curve-fitting capabilities to include Polynomial, Power, and Exponential models, leveraging the tcplfit2 R package.
• Modified invitrodb (beta version v4.0) to support new modeling parameters and updated model selection logic.
• Implemented continuous hit calling logic and removed redundant endpoints through bidirectional fitting.

Main Results:

• The updated tcpl and invitrodb (v4.0) accommodate a wider array of curve-fitting models.
• Hit call and potency estimates remained largely consistent between invitrodb versions 3.5 and 4.0.
• The enhancements facilitate standardized, reproducible curve-fitting and data management for in vitro toxicology data.

Conclusions:

• Tcpl and invitrodb provide a robust standard for managing and analyzing diverse in vitro assay data.
• These updates enhance data comparability across different tiers within the US Environmental Protection Agency CompTox Blueprint.
• The improved software and database support broader applications of ToxCast data in toxicology and chemical risk assessment.