Identification of gene expression signature for drought stress response in barley (Hordeum vulgare L.) using machine learning approach

Abstract

Barley (Hordeum vulgare L.) is an important cereal crop, playing a pivotal role in global agriculture and food systems. Drought has a significant impact on barley growth and yield productivity. In the current study, core drought stress responsive genes were investigated using an integrative approach. First, we determined the core differentially expressed genes (DEGs) in multiple RNA-seq experiments using a p-value combination approach. Then, machine learning approaches including four weighting algorithms were harnessed for prioritization and determination of signature genes. Moreover, predictive models were optimized using tree induction and naive Bayes algorithms. Finally, the functional importance of the core DEGs and signature genes and pathways were dissected using gene ontology, KEGG enrichment, and protein-protein interaction network analysis. Results showed that the core DEGs participate in carbon metabolism, biosynthesis of secondary metabolites, glyoxylate and dicarboxylate metabolism, carbon fixation, biosynthesis and degradation of amino acids, glycolysis/gluconeogenesis, pyruvate metabolism, starch and sucrose metabolism, glycerolipid metabolism, beta-alanine metabolism, ascorbate and aldarate metabolism, taurine and hypotaurine metabolism. Notably, the C4.5 algorithm, boasting a remarkable 100 % accuracy, pinpointed two genes of particular importance including HORVU.MOREX.R3.1HG0063740, encoding the endo-1, 3–1, 4-beta-D-glucanase, and HORVU.MOREX.R3.1HG0083720, which encodes the bifunctional inhibitor/lipid-transfer protein. This comprehensive analysis contributes significantly to understanding of the core drought responsive genes and pathways. Moreover, these findings lay the groundwork for further research aimed at developing drought-resistant barley varieties and utilizing predictive models in field screening programs.