Legume Genomics and Genetics 2026, Vol.17, No.1, 32-48 http://cropscipublisher.com/index.php/lgg 44 network analysis to pinpoint genes involved in hormone signaling, antioxidant defense, secondary metabolism, and cell wall remodeling (Shahriari et al., 2022; Tang et al., 2023). At the germination stage, integration of RTM-GWAS with root transcriptomes led to the identification of 58 QTLs and 22 candidate genes, including transcription factors and transporters whose drought-induced expression patterns align with tolerant phenotypes. Similar transcriptome-driven surveys across seed developmental stages have revealed stage-specific DEGs related to heat shock proteins, LEA proteins, and regulatory TFs, providing novel candidates for protecting yield under reproductive-stage drought. These studies underscore the value of transcriptomics for cataloguing stress-induced genes and connecting them to specific tissues, developmental windows, and physiological traits. New computational pipelines are significantly improving the efficiency and precision of this gene mining. An integrative data-driven feature-engineering (DFE) framework aggregates cloud-based transcriptomic, genomic, and non-omics datasets and applies robust feature prioritization to identify key drought-tolerant genes (DT genes) while reducing noise and false positives relative to traditional WGCNA (Kao et al., 2025). Multi-omics-based alternative splicing prediction in wild soybean has further expanded the candidate space by revealing 139 genes co-expressed at transcript and protein levels and subject to drought-induced alternative splicing, with isoform-specific regulation of genes such as FT1, CCR1L, and RPL18 linked to drought adaptation (Kim et al., 2024). Joint transcriptome-proteome and transcriptome-metabolome analyses in roots and seedlings have highlighted phenylpropanoid, flavonoid, and TCA-cycle pathways as core drought-resistance modules and nominated both structural genes and pathway regulators as breeding targets (Zhao et al., 2021; Wang et al., 2022). Together, these advances are transforming transcriptomics from a descriptive tool into a discovery engine that yields prioritized, experimentally tractable drought-resistance gene resources. 7.2 Molecular markers and genomic selection in breeding The increasing availability of high-density SNP data and transcriptome-defined candidate genes has accelerated the development of molecular markers for drought tolerance in soybean. GWAS using tens of thousands to millions of SNPs, coupled with PEG-based germination assays and multi-trait drought indices, have detected numerous QTLs controlling germination rate, root traits, and seedling vigor under water deficit (Kong et al., 2025). For instance, RTM-GWAS with 95,043 SNPs identified 58 QTLs at germination, including 10 large-effect loci; by intersecting these regions with drought-responsive DEGs and co-expression modules, 22 high-confidence candidate genes were defined as valuable genetic resources for breeding (Kong et al., 2025). A separate GWAS of 264 Chinese accessions with 2,597,425 SNPs detected 92 significant SNPs and nine candidate genes, and led to the development of two Kompetitive Allele Specific PCR (KASP) markers tightly linked to drought tolerance at germination, providing low-cost, high-throughput tools for marker-assisted selection (MAS) (Jia et al., 2024). Traditional QTL mapping using high-density SLAF-seq maps has also identified genomic regions controlling plant height and seed weight per plant under drought, with several major and common QTLs proposed for deployment in MAS (Ren et al., 2020). Transcriptomics enhances these marker efforts by providing functional context and enabling candidate-gene-based markers and genomic prediction models. Expression profiling within QTL intervals helps prioritize functional genes, such as those regulating ion homeostasis, plasma membrane ATPase activity, or heat-shock protein synthesis, which can then be converted into diagnostic markers for drought tolerance (Park et al., 2025). Integrating expression-based co-expression modules and hub genes into genomic selection frameworks allows breeders to weight markers near regulatory hubs or pathway bottlenecks more heavily, potentially increasing prediction accuracy for complex traits like drought tolerance (Shahriari et al., 2022). As sequencing costs fall, genomic selection models trained on whole-genome SNP data, multi-environment drought phenotypes, and transcriptomic signatures from key tissues and stages are expected to become standard tools. These models can accelerate the identification of superior lines and support pyramiding of multiple drought-tolerance loci while maintaining yield and quality under variable climates (Valliyodan et al., 2016).
RkJQdWJsaXNoZXIy MjQ4ODYzNA==