Transcriptome-Wide Co-Expression Network Analysis Identifies Candidate Genes Associated with Seed Protein Content in Soybean

The transcriptional processes associated with final seed protein content are not fully characterized in soybean, particularly in progeny derived from crosses with wild soybean. To address this knowledge gap, high-protein (HP) and low-protein (LP) BC1F3 progeny were compared at 25, 35, and 45 days after flowering (D25, D35, and D45) using seed-composition measurements, RNA sequencing (RNA-seq), differential expression analysis, temporal expression clustering, functional enrichment, and weighted gene co-expression network analysis (WGCNA). The HP and LP phenotypes became most distinct late in seed development: at D45, the protein content was 9.0% higher in HP, whereas the oil content was 7.3% lower. The number of differentially expressed genes (DEGs) between HP and LP increased from 3386 at D25 to 7269 at D45, and 18,394 genes showed significant Group × Time interactions. Six temporal clusters revealed stage-dependent expression patterns enriched for cell division, photosynthesis, reserve-related functions, translation, protein processing, and redox-associated processes. WGCNA identified pink and salmon modules positively correlated with protein content and a brown module negatively correlated with protein content. Integration of stage-specific DEGs, temporal interactions, and network metrics prioritized candidates including Glyma.11G063900.1, Glyma.06G030400.1, Glyma.07G038500.2, and Glyma.03G065500.2. Modules correlated with oil content were interpreted primarily as developmental modules because the oil content was nearly collinear with the developmental stage. These results provide a stage-resolved framework for prioritizing candidate genes for functional validation and breeding aimed at improving soybean seed protein content.