Variability in drought gene expression datasets highlight the need for community standardization.

Area of Science:

• Plant science
• Molecular biology
• Genomics

Background:

• Physiologically relevant drought stress application is inconsistent across studies.
• Heterogeneity in experimental design, growth conditions, and sampling complicates comparisons of plant water deficit research.
• Many gene expression studies lack crucial physiological data to validate stress conditions.

Purpose of the Study:

• To re-analyze hundreds of drought gene expression experiments to quantify inter-study variability.
• To identify factors contributing to the lack of comparability in plant drought stress studies.
• To develop a quality control method for assessing the physiological relevance of drought stress in gene expression datasets.

Main Methods:

• Re-analysis of hundreds of plant drought gene expression experiments.
• Quantification of variability across studies, controlling for genotype, environment, and drought severity.
• Development of supervised learning classifiers to predict physiologically relevant drought stress from RNA-sequencing data.

Main Results:

• Plant drought stress studies exhibit significant and surprising incomparability, even when controlling for key variables.
• A substantial number of studies, particularly in Arabidopsis, lack adequate phenotypic and physiological data.
• Supervised learning classifiers were developed that can accurately predict physiologically relevant drought stress in RNA-sequencing samples.

Conclusions:

• Community-wide standardization of experimental protocols is essential for reproducible plant drought stress research.
• Paired physiological data is critical for quantifying stress severity and ensuring the reliability of gene expression analyses.
• The developed classifiers can serve as a valuable quality control tool for future plant drought stress studies.