gsGNA: A Universal Pipeline for Mining Plant Stress-Resistance Regulators

Background Under global warming, important food crops such as rice are severely impacted by abiotic stresses. However, few studies have explored the mechanisms underlying abiotic stress responses from a systems biology perspective, limiting the identification of breeding targets for developing novel rice varieties with enhanced stress tolerance. Results Utilizing rice RNA-seq data under aridity, alkaline, and cold stress conditions, we proposed gsGNA (General Stress-resistance Gene Regulatory Network), a universal computational pipeline designed to identify plant stress-resistance regulators. This pipeline employs six ensemble strategies to integrate the predictive outputs of six machine learning and deep learning algorithms, thereby systematically constructing high-confidence gene regulatory networks (GRNs) of rice under abiotic stress. Comparative analysis demonstrated that the XGBoost ensemble method consistently exhibited stable predictive performance across all three stress conditions. Subsequent network analysis identified 20 core transcription factors (TFs) that are shared among the three stresses. Functional enrichment analysis revealed that these TFs and their target genes cooperatively regulate key biological pathways, including hormone signal transduction, the MAPK signaling pathway, and primary metabolism. Network module analysis indicated that GRNs under different stresses are highly conserved in their overall architecture but exhibit significant divergence at the functional module level, reflecting an organizational pattern characterized by a "conserved core and divergent periphery". Additionally, through a reverse screening strategy based on core pathway target genes, 45 potentially novel regulators were predicted using the GRNs, providing new clues for further dissecting rice abiotic stress responses. Conclusions By applying the proposed gsGNA pipeline, we systematically mapped the GRN landscape of rice in response to common abiotic stresses, elucidated its network organizational patterns, and identified a set of candidate key TFs. These findings provide a theoretical foundation and a valuable target repository for rice stress-tolerance breeding.