Classification of biodegradable materials using QSAR modelling with uncertainty estimation.

Area of Science:

• Environmental Chemistry
• Computational Chemistry
• Toxicology

Background:

• Determining chemical biodegradability is crucial for ecological and economic reasons.
• Quantitative structure-activity relation (QSAR) models offer a cost-effective alternative to traditional testing.
• Existing QSAR models often lack detailed uncertainty estimates for individual predictions, limiting their regulatory utility.

Purpose of the Study:

• To develop and apply a QSAR model that provides uncertainty estimates for biodegradability predictions.
• To enhance the trustworthiness and regulatory applicability of computational models for chemical assessment.
• To demonstrate a method for assigning confidence intervals to individual predictions.

Main Methods:

• Utilized a partial least squares discriminant analysis (PLS-DA) model for chemical separation.
• Employed bootstrapping techniques to estimate prediction uncertainty.
• Applied the developed QSAR model to predict biodegradability for a set of substances from a public dataset.

Main Results:

• The QSAR model successfully incorporated uncertainty estimation into biodegradability predictions.
• Bootstrapping provided a means to assign confidence intervals to individual predictions.
• The uncertainty estimates allow for a more thorough model assessment than traditional statistical methods.

Conclusions:

• The developed QSAR model with uncertainty estimation significantly improves the assessment of prediction reliability.
• This approach enhances the utility of computational models for regulatory decision-making.
• The method for calculating uncertainty highlights specific areas where additional experimental testing is most beneficial.