Predicting persistence in the sediment compartment with a new automatic software based on the k-Nearest Neighbor...

Area of Science:

• Environmental chemistry
• Computational toxicology
• Predictive modeling

Background:

• Assessing chemical persistence is crucial for environmental and human health protection.
• Regulatory frameworks mandate persistence evaluations for manufactured substances.
• In silico methods offer an alternative to laboratory testing for predicting persistence.

Purpose of the Study:

• To develop and validate a novel computational program for predicting chemical persistence in the environment.
• To utilize the k-Nearest Neighbor (k-NN) algorithm for property prediction based on chemical similarity.
• To specifically model persistence in the sediment compartment.

Main Methods:

• Development of a k-Nearest Neighbor (k-NN) modeling program.
• Compilation and curation of sediment half-life (HL) data for 297 organic compounds.
• Division of the dataset into four experimental classes for model training and testing.
• Evaluation of model performance with training and test set accuracies ranging from 0.90 to 0.96.

Main Results:

• Several k-NN models for predicting sediment persistence demonstrated satisfactory performance.
• Selected model achieved high accuracy on both training and test datasets (0.90-0.96).
• The developed model provides reliable in silico predictions for chemical persistence.

Conclusions:

• The new k-NN model is a valuable in silico tool for rapid and cost-effective screening of chemical persistence.
• This predictive model will be integrated into the VEGA software platform for wider accessibility.
• The tool supports environmental risk assessment and regulatory compliance by predicting substance degradation.