Mechanism and Prediction of Gray Jujube Fruit Quality Using Explainable ANN.

Area of Science:

• Agricultural Science
• Plant Physiology
• Computational Biology

Background:

• Gray jujube (Ziziphus jujuba) is a vital economic fruit crop in Xinjiang, China.
• Fruit quality is influenced by complex interactions between tree architecture, physiology, and environment.

Purpose of the Study:

• Develop an interpretable artificial neural network model to predict key quality parameters of gray jujube.
• Identify key structural and physiological indicators influencing fruit quality.

Main Methods:

• Field experiments over two years.
• Development and optimization of an artificial neural network model using Bayesian optimization.
• Integration of 13 structural and physiological indicators.

Main Results:

• Achieved high prediction accuracy (R² = 0.89-0.98) for vitamin C, soluble sugar, titratable acid, and sugar-acid ratio.
• Identified shoot elongation and SPAD values as crucial for vitamin C accumulation.
• Determined optimal leaf inclination angles and light transmittance for sugar accumulation.
• Found that high net photosynthetic rates reduce organic acid content.

Conclusions:

• The developed model effectively predicts gray jujube quality and provides mechanistic insights.
• Findings support precision cultivation strategies for enhancing fruit quality.
• The integrated framework offers a comprehensive approach for managing crop quality.