Integrative cross-platform transcriptomic analysis coupled with machine learning identifies candidate biomarkers of abiotic stress resilience in Arabidopsis thaliana.

backgroundEnvironmental stressors impose significant constraints on plant growth and productivity, creating an imperative to elucidate the molecular mechanisms underlying stress tolerance. A major challenge in this domain is integrating large and heterogeneous transcriptomic datasets generated using diverse technological platforms. To address this challenge, we applied a pooled, batch-corrected integrative transcriptomic framework to harmonize microarray and RNA-seq datasets and identify core transcriptional responses in Arabidopsis thaliana exposed to abiotic stress.

resultsUsing variance-stabilizing transformation for RNA-seq normalization prior to integration, our pipeline enabled robust detection of 102 consensus differentially expressed genes (84 upregulated and 18 downregulated). Functional enrichment analysis revealed strong associations with abiotic stimulus perception, protein stabilization, and cellular stress adaptation. Network topology analysis highlighted At5g05410 (DREB2A) and At2g26150 (HSFA2) as central regulatory hubs, suggesting coordinated crosstalk between dehydration-responsive and heat shock pathways. Machine-learning-based prioritization identified six top candidate genes (At1g53540, At4g18280, At5g12030, At1g09750, At2g40610, and At2g03750), with XGBoost achieving the highest classification performance (AUC = 0.98) under internal cross-validation. This performance should be interpreted in light of potential class imbalance, prior DEG-based feature preselection, and the absence of independent external validation. RT-qPCR validation across seven abiotic stresses demonstrated distinct stress-adaptive transcriptional signatures. The small heat shock proteins At1g53540 (HSP17.6 C) and At5g12030 were strongly induced under heat, salt, and drought conditions, while the cell wall-associated genes At2g40610 and At2g03750 were persistently downregulated, consistent with a growth–defense trade-off. Promoter motif analysis further supported these regulatory patterns, revealing enrichment of ABRE and HSE elements in induced genes and ARE motifs in genes repressed under nutrient-limited conditions.

conclusionThis integrative transcriptomic analysis combined with experimental validation identifies high-confidence genes and regulatory modules associated with abiotic stress adaptation in Arabidopsis thaliana. The findings offer mechanistic insights into stress-responsive transcriptional regulation and highlight candidate genes and regulatory modules that warrant further functional validation before consideration for crop-improvement applications.