Using Network-Based Machine Learning to Predict Transcription Factors Involved in Drought Resistance

Area of Science:

• Plant biology
• Genomics
• Bioinformatics

Background:

• Gene regulatory networks (GRNs) are crucial for plant stress responses.
• Identifying key genes within complex GRNs for specific traits like drought resistance is difficult.

Purpose of the Study:

• To develop an interactive web platform, GRAiN, for analyzing rice (Oryza sativa) gene regulatory networks.
• To create a machine learning framework for prioritizing genes involved in drought stress response in rice.

Main Methods:

• Constructed the Gene Regulation and Association Network (GRAiN) using multiple network inference algorithms on public gene expression data.
• Developed a supervised machine learning model to assign a 'drought score' to 2160 transcription factors (TFs).
• Validated predicted TF functions using in vitro protein-DNA binding assays and loss-of-function mutant analysis.

Main Results:

• The GRAiN platform enables the study of functional relationships between TFs and genetic modules in abiotic stress responses.
• The machine learning framework successfully converted network data into a prioritized list of drought-responsive TFs.
• Highest-scoring TFs revealed key functional, structural, and evolutionary aspects of rice drought resistance.
• The function of OsbHLH148 was accurately predicted and experimentally validated.

Conclusions:

• GRAiN and the machine learning strategy provide a powerful approach for identifying key regulatory genes in plants.
• This methodology can accelerate the genetic engineering of desirable traits, such as drought resistance, in rice and other crops.