The stochastic regression analysis as a tool in ecotoxicological QSAR studies

Area of Science:

• Environmental Chemistry
• Ecotoxicology
• Quantitative Structure-Activity Relationships (QSAR)

Background:

• Benzene derivatives are common environmental pollutants.
• Understanding their ecotoxicological impact on aquatic life, such as Pimephales promelas (fathead minnow), is crucial.
• Quantitative Structure-Activity Relationships (QSAR) are vital for predicting chemical toxicity.

Purpose of the Study:

• To investigate the structure-activity relationships of 50 benzene derivatives.
• To develop a predictive model for the acute toxicity (96-h LC50) of these compounds on Pimephales promelas.
• To compare the efficacy of Correspondence Factor Analysis (CFA) with Principal Component Analysis (PCA) in QSAR modeling.

Main Methods:

• Utilized Correspondence Factor Analysis (CFA) and linear regression analysis.
• Derived five new independent variables from nine molecular descriptors (C, H, O, N, Br, Cl, NO2, OH, NH2).
• Employed a stepwise regression analysis to build the QSAR model and compared it with a PCA-based model.

Main Results:

• A QSAR model based on CFA (log 1/C = -0.727F1 + 1.248F3 + 4.052) showed high relevance (r = 0.918, s = 0.270).
• The CFA-derived model demonstrated superior predictive power for fathead minnow toxicity compared to the PCA model (r = 0.737, s = 0.460).
• The study highlights the effectiveness of stochastic regression analysis in ecotoxicological assessments.

Conclusions:

• Correspondence Factor Analysis provides a more relevant approach for modeling the ecotoxicological behavior of benzene derivatives.
• The identified QSAR model accurately predicts the acute toxicity of aromatic compounds on Pimephales promelas.
• This statistical approach enhances the understanding of chemical impacts on aquatic ecosystems.