A SAR-Based Classification Model for Assessing Pesticide Toxicity to Apis mellifera

Area of Science:

• Environmental toxicology
• Computational chemistry
• Apiculture science

Background:

• Pollinator populations are declining due to pesticides, parasites, and climate change.
• Protecting pollinators is crucial for ecosystem stability and food security.
• Current pesticide risk assessments are costly and time-consuming.

Purpose of the Study:

• To develop a computational model for predicting the oral acute toxicity of pesticides in honey bees (Apis mellifera).
• To utilize Structure-Activity Relationship (SAR) models for efficient pesticide screening.
• To support regulatory efforts in pesticide risk assessment.

Main Methods:

• Developed a classification model using a dataset of 357 compounds.
• Employed structural alerts to predict oral acute toxicity in Apis mellifera.
• Validated the model using training and test sets.

Main Results:

• The model achieved a high Matthews Correlation Coefficient (0.82) in the training set.
• The test set showed a moderate decay (0.56) possibly due to applicability domain limits.
• High balanced accuracy (0.80) and sensitivity (0.79) were observed in the test set.

Conclusions:

• The developed SAR model is a reliable tool for the toxicological screening of pesticides.
• Computational approaches offer an efficient alternative to traditional methods for assessing pollinator toxicity.
• This model can aid in prioritizing chemicals for further testing and safeguarding pollinators.