Phenotype prediction in plants is improved by integrating large-scale transcriptomic datasets.

Area of Science:

• Plant biology
• Genomics
• Bioinformatics

Background:

• Understanding dynamic gene expression is crucial for plant biological processes.
• Transcriptomic data offers insights into gene activity across various plant conditions.

Purpose of the Study:

• To investigate if highly variable genes (HVGs) can accurately identify plant phenotypes.
• To develop a machine learning (ML) framework for phenotype prediction using gene expression data.

Main Methods:

• Utilized a large dataset of 21,5612 maize (Zea mays L.) bulk RNA sequencing samples.
• Developed and applied machine learning models to predict plant phenotypes based on HVG expression levels.
• Validated the approach in rice (Oryza sativa L.) to assess cross-species generalizability.

Main Results:

• Machine learning models achieved high accuracy in predicting maize phenotypes (tissue type, developmental stage, cultivar, stress) using only HVGs.
• Identified several key functional genes associated with distinct plant phenotypes through ML analysis.
• Demonstrated that while ML models show cross-species potential, direct transferability between maize and rice is limited due to gene expression divergence.

Conclusions:

• A robust ML framework for phenotype prediction from gene expression profiles has been established.
• The findings support the use of HVGs for accurate plant phenotype identification.
• This research contributes to the advancement of precision crop management in agriculture.