Uncovering waterlogging-responsive genes in cucumber through machine learning and differential gene correlation...

Area of Science:

• Plant Science
• Genomics
• Bioinformatics

Background:

• Climate change is increasing waterlogging events, posing a significant threat to cucumber production.
• Understanding cucumber's genetic response to waterlogging is crucial for developing resilient crop varieties.

Purpose of the Study:

• To identify key genes involved in cucumber's response to waterlogging stress.
• To validate the reliability of identified genes using multiple feature selection and classification methods.
• To provide insights into the molecular mechanisms underlying cucumber waterlogging tolerance.

Main Methods:

• Analysis of three public RNA-seq datasets (36 samples) using various feature selection algorithms (Uncertainty, Relief, SVM, Correlation, LASSO).
• Differential gene correlation analysis (DGCA) to identify changes in gene interaction networks under stress.
• Validation of selected genes using Random Forest (RF) classifier and SHapley Additive exPlanations (SHAP) values.
• Cross-validation with Boruta algorithm to confirm gene selection robustness.

Main Results:

• Multiple feature selection methods identified significant genes related to waterlogging stress.
• DGCA revealed significant changes in the correlation of 13 genes under waterlogging.
• The Random Forest model achieved 100% accuracy, with LOC101209599, LOC101217277, and LOC101216320 identified as highly impactful.
• Eight genes were consistently identified across multiple feature selection algorithms, including LASSO, DGCA, and Boruta.

Conclusions:

• Key genes (e.g., LOC101209599, LOC101217277, LOC101216320) are critical for cucumber waterlogging tolerance.
• These genes are involved in vital pathways including stress tolerance, root development, and calcium signaling.
• The identified genes and pathways offer valuable targets for breeding cucumber varieties with enhanced waterlogging resistance.