An artificial neural network-based deep learning model to predict combined stress impact and interaction in plants.

Area of Science:

• Plant Science
• Computational Biology
• Agronomy

Background:

• Plants face combined abiotic and biotic stresses, posing significant threats to crop yield.
• Experimental data for numerous stress combinations is scarce, hindering comprehensive understanding.
• Existing literature data is underutilized for predicting complex plant stress responses.

Purpose of the Study:

• To develop a computational tool for predicting the impact of combined stresses on plants.
• To overcome limitations in experimental data generation for plant stress research.
• To provide a resource for understanding multivariate and complex combined stress datasets.

Main Methods:

• Literature data on plant stress combinations was gathered from public databases.
• A composite artificial neural network (ANN)-based deep learning model was developed.
• Machine learning algorithms were employed for multi-target classification and regression.

Main Results:

• The ANN model accurately predicted the impact of stress interactions on plant morphological parameters (76.33% accuracy).
• The model quantified percentage changes in affected morphological parameters.
• Predicted yield reductions were validated in rice under combined drought and heat stress.

Conclusions:

• The developed ANN model is a valuable resource for plant researchers.
• The tool facilitates understanding of complex plant stress interactions.
• This computational approach accelerates the generation of biological inferences on plant stress responses.