Integrative Analysis of Abiotic Stress-Responsive Genes in Soybean Using Differential Gene Expression and Validation With Machine Learning.

The increasing frequency and intensity of climate-associated abiotic stresses highlight the need to better understand soybean stress-response mechanisms. In this study, transcriptomic data from 112 soybean samples exposed to multiple abiotic stress conditions, including drought, salinity, and heavy-metal stress, were integrated using a multi-layer computational framework. By combining differential expression analysis, Differential Gene Correlation Analysis (DGCA), Random Forest, XGBoost, functional enrichment, and network-based interpretation, the analysis prioritized 37 candidate genes associated with abiotic stress responses across heterogeneous experimental contexts. Functional enrichment analyses indicated that these candidates were associated with amino acid biosynthesis, energy metabolism, carbon metabolism, glycolysis, secondary metabolite biosynthesis, and ROS-related processes. DGCA further suggested stress-associated changes in gene-gene correlation patterns, supporting the presence of condition-dependent transcript coordination rather than direct causal regulation. Among the prioritized candidates, GLYMA_16G207700, GLYMA_16G204600, and GLYMA_16G214500 emerged as high-priority genes based on multi-layer analytical support, chromosomal localization, and comparative genomic evidence. Their orthology or syntenic relationships with loci in