Deep learning uncovers distinct behavior of rice network to pathogens response

Area of Science:

• Plant pathology
• Computational biology
• Genomics

Background:

• Rice is susceptible to various pathogens, including bacterial *Xanthomonas oryzae* (*Xoo*) and fungal *Magnaporthe oryzae*.
• Previous research has extensively studied these pathogens but often lacked holistic network analysis.

Purpose of the Study:

• To design a deep learning-based rice network model (DLNet) for analyzing quantitative differences in rice immune responses.
• To explore distinct rice network architectures under biotic stress.

Main Methods:

• Development of a deep learning-based rice network model (DLNet).
• Validation of DLNet on rice in response to biotic stresses, comparing its performance against other machine learning methods.
• Analysis of network architecture differences between rice- *M. oryzae* and rice- *Xoo* interactions.

Main Results:

• DLNet demonstrated superior performance compared to other machine learning methods in validating rice responses to biotic stresses.
• Identified compactness in the rice PTI (PAMP-triggered immunity) network and independent modules in the rice ETI (effector-triggered immunity) network.
• Observed more independent network modules and less structural disorder in the rice- *M. oryzae* model versus the rice- *Xoo* model.

Conclusions:

• The study highlights distinct rice immune network architectures shaped by different pathogens.
• Findings suggest unique adaptation strategies employed by rice to evade effectors from *M. oryzae* compared to *Xoo*.
• The DLNet model provides a powerful tool for dissecting plant immune responses at a network level.