Related Experiment Video
Updated: May 11, 2026
12:43
Electroantennographic Bioassay as a Screening Tool for Host Plant Volatiles
Published on: May 6, 2012
14.1K
Accelerating VOC-OBP interaction screening via machine learning and molecular docking: Towards semiochemicals
Xaviera López-Cortés1, Nicolás Fernández2, Gabriel Lara2
1Department of Computer Sciences and Industries, Universidad Católica del Maule, Talca 3466706, Chile; Centro de Innovación en Ingeniería Aplicada (CIIA), Universidad Católica del Maule, Talca 3466706, Chile.
Computational Biology and Chemistry
|November 22, 2025
Summary
This study uses machine learning to predict moth odorant-binding protein (OBP) interactions with volatile organic compounds (VOCs). The LightGBM model achieved the best predictive accuracy, accelerating molecular screening for pest management.
Area of Science:
- Computational chemistry
- Bioinformatics
- Machine learning
Background:
- Odorant-binding proteins (OBPs) are crucial for insect olfaction.
- Predicting binding affinity between volatile organic compounds (VOCs) and OBPs is vital for developing semiochemical-based pest management strategies.
- Current experimental methods for screening molecular interactions are time-consuming and labor-intensive.
Purpose of the Study:
- To develop and evaluate a machine learning-based approach for predicting the binding affinity between VOCs and moth OBPs.
- To accelerate the process of molecular screening for potential semiochemicals.
- To reduce the experimental workload in identifying molecules for pest control.
Main Methods:
- A diverse set of regression models were evaluated, including ensemble methods (LightGBM, XGBoost, Gradient Boosting, Random Forest), Support Vector Regressor, Convolutional Neural Network, and Bayesian linear models.
- Model performance was assessed using metrics such as the coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE).
Main Results:
- The LightGBM Regressor demonstrated superior performance with an R2 of 0.7101, RMSE of 0.2979, and MAE of 0.2200.
- Ensemble-based boosting algorithms, like LightGBM, effectively captured complex, non-linear relationships in the binding affinity data.
- Linear models, such as the Bayesian Ridge Regressor, showed limited predictive accuracy, indicating the complexity of the interactions.
Conclusions:
- Machine learning, particularly ensemble methods, offers a powerful and efficient approach for predicting VOC-OBP binding affinities.
- The developed models can significantly accelerate molecular screening, aiding in the discovery of semiochemicals for integrated pest management.
- Future work will focus on expanding datasets and refining model architectures to improve prediction quality and generalizability.
