Novel Bayesian Method to Derive Final Adjusted Values of Physicochemical Properties: Application to 74 Compounds

Area of Science:

• Environmental Chemistry
• Computational Chemistry
• Toxicology

Background:

• Accurate physicochemical properties are crucial for assessing human and environmental risks of semivolatile organic compounds (SVOCs).
• Existing in silico estimation methods and literature-reported measurements often lack consistent accuracy and reliability.
• Final Adjusted Values (FAVs) are needed to harmonize data and minimize uncertainty in risk assessments.

Purpose of the Study:

• To develop and validate a novel Bayesian workflow for estimating Final Adjusted Values (FAVs) of physicochemical properties for SVOCs.
• To compare the performance of various in silico estimation methods and measurement techniques.
• To generate a set of recommended FAVs (FAV^Rs) for a diverse range of SVOCs.

Main Methods:

• A novel Bayesian approach was developed to combine direct and indirect measurement data with in silico estimations.
• The workflow was applied to 74 compounds across nine classes.
• Performance of in silico models (OPERA, COSMOtherm, EPI Suite, SPARC, pp-LFER) and measurement methods was evaluated.

Main Results:

• In silico methods showed significant variability in accuracy, performing poorly for larger, more polar compounds.
• COSMOtherm and OPERA demonstrated generally good performance with low bias, but no single in silico method was universally superior.
• Indirect measurement methods provided highly accurate and precise estimates compared to direct measurements.
• The Bayesian method successfully harmonized diverse data sources without introducing bias.

Conclusions:

• The developed Bayesian workflow effectively harmonizes physicochemical property data for SVOCs.
• Recommended FAVs (FAV^Rs) are provided for 74 compounds, suitable for hazard and risk screening.
• The workflow is recommended for generating FAV^Rs for additional SVOCs, enhancing risk assessment reliability.