Scholarlink

<p>Drought and salinity are major environmental constraints that can reduce crop productivity, and understanding the regulatory networks involved in plant adaptation to these overlapping osmotic stresses may contribute to efforts aimed at improving climate resilience in crops. In this study, we applied an integrated systems biology framework combining cross-study transcriptomic integration of multiple independent datasets, Weighted Gene Co-expression Network Analysis (WGCNA), deep learning-based autoencoder modeling, and RT-qPCR validation to investigate transcriptional responses in <i>Arabidopsis thaliana</i>. Eight independent transcriptomic datasets comprising 40 samples (20 stress-treated and 20 controls) were analyzed, resulting in the identification of 678 differentially expressed genes (450 upregulated and 228 downregulated). Functional enrichment analysis indicated that upregulated genes were associated with stress-responsive pathways and transcription factor families such as NAC and MYB, whereas downregulated genes were enriched in processes related to cell wall organization and growth. Network analysis identified three candidate co-expression modules (Black, Salmon, and Turquoise) that were significantly correlated with stress traits. The deep autoencoder model showed high classification performance for the selected stress-responsive genes (Accuracy = 93%, AUC = 0.993) and supported feature compression and prioritization of candidate regulators. The cross-study transcriptomic results, network topology, and deep learning outputs collectively prioritized seven candidate hub genes (<i>At2g33380</i>, <i>At2g41990</i>, <i>At2g42790, At2g02390, At2g35940, At2g38310</i>, and <i>At5g64620</i>). From these, three candidates with comparatively limited experimental documentation under these stress conditions–<i>At2g42790 (CSY3)</i>, <i>At5g64620 (C/VIF2)</i>, and <i>At2g41990 (T6D20.12)</i>–were selected for RT-qPCR validation. Subsequent RT-qPCR analysis showed significant upregulation of <i>At2g42790</i> (<i>CSY3</i>) and downregulation of <i>At5g64620</i> (<i>C/VIF2</i>) and <i>At2g41990</i> (<i>T6D20.12</i>), providing additional support for the in-silico predictions. Collectively, these findings are consistent with a coordinated transcriptional response under osmotic stress, characterized by increased representation of defense-related pathways and reduced representation of growth-associated processes. This study suggests that integrating deep learning with network biology may be a useful framework for identifying candidate genes for further research on abiotic stress tolerance in crops.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Systems biology and deep learning reveal master regulators of convergent drought and salinity stress in Arabidopsis thaliana validated by RT-qPCR

  • 期刊论文

  • 发表日期：

    2026年8月3日

查看全文
  • Maryam Mehdizadeh Hakkak,
  • Masoud Tohidfar

摘要

Drought and salinity are major environmental constraints that can reduce crop productivity, and understanding the regulatory networks involved in plant adaptation to these overlapping osmotic stresses may contribute to efforts aimed at improving climate resilience in crops. In this study, we applied an integrated systems biology framework combining cross-study transcriptomic integration of multiple independent datasets, Weighted Gene Co-expression Network Analysis (WGCNA), deep learning-based autoencoder modeling, and RT-qPCR validation to investigate transcriptional responses in Arabidopsis thaliana. Eight independent transcriptomic datasets comprising 40 samples (20 stress-treated and 20 controls) were analyzed, resulting in the identification of 678 differentially expressed genes (450 upregulated and 228 downregulated). Functional enrichment analysis indicated that upregulated genes were associated with stress-responsive pathways and transcription factor families such as NAC and MYB, whereas downregulated genes were enriched in processes related to cell wall organization and growth. Network analysis identified three candidate co-expression modules (Black, Salmon, and Turquoise) that were significantly correlated with stress traits. The deep autoencoder model showed high classification performance for the selected stress-responsive genes (Accuracy = 93%, AUC = 0.993) and supported feature compression and prioritization of candidate regulators. The cross-study transcriptomic results, network topology, and deep learning outputs collectively prioritized seven candidate hub genes (At2g33380, At2g41990, At2g42790, At2g02390, At2g35940, At2g38310, and At5g64620). From these, three candidates with comparatively limited experimental documentation under these stress conditions–At2g42790 (CSY3), At5g64620 (C/VIF2), and At2g41990 (T6D20.12)–were selected for RT-qPCR validation. Subsequent RT-qPCR analysis showed significant upregulation of At2g42790 (CSY3) and downregulation of At5g64620 (C/VIF2) and At2g41990 (T6D20.12), providing additional support for the in-silico predictions. Collectively, these findings are consistent with a coordinated transcriptional response under osmotic stress, characterized by increased representation of defense-related pathways and reduced representation of growth-associated processes. This study suggests that integrating deep learning with network biology may be a useful framework for identifying candidate genes for further research on abiotic stress tolerance in crops.