Identifying Synergistic Components of Botanical Fungicide Formulations Using Interpretable Graph Neural Networks.

Area of Science:

• Agricultural Science
• Computational Biology
• Biochemistry

Background:

• Botanical formulations offer environmentally friendly biopesticide potential.
• Enhancing biopesticide efficacy with permeation enhancers is crucial but challenging due to complex chemical interactions.
• Discovering synergistic combinations requires efficient predictive methods.

Purpose of the Study:

• To develop a novel deep learning method for predicting synergy between botanical products and permeation enhancers.
• To enable efficient identification of effective biopesticide formulations.
• To provide insights into the mechanisms of synergistic interactions.

Main Methods:

• A deep learning model utilizing weighted component feature vectors to represent chemical mixtures.
• Handling variable numbers of components and interpreting individual contributions to synergy.
• Ensemble interpretation methods to elucidate underlying synergistic mechanisms.

Main Results:

• The deep learning method accurately predicts synergistic combinations of botanical biopesticides and permeation enhancers.
• Wet-lab validation confirmed the discovery of novel and effective biopesticide formulations.
• The approach demonstrated generalizability beyond biopesticide applications.

Conclusions:

• The proposed deep learning method significantly enhances the discovery of synergistic chemical mixtures.
• This approach facilitates the development of effective and environmentally sustainable biopesticides.
• The method is adaptable for predicting synergistic effects in various chemical compound mixture applications.