Machine learning-based transcriptome mining to discover key genes for density stress in sweet corn

Abstract

Sweet corn stands as a crucial staple in the food industry, offering consumers a nutritious and diverse option. However, understanding its response to density stress remains pivotal for enhancing its resilience and productivity. We employed Weighted Gene Co-expression Network Analysis (WGCNA), differential gene expression analysis, and Least Absolute Shrinkage and Selection Operator (LASSO) regression to dissect its molecular mechanisms. Four key genes (GRMZM2G129246, GRMZM2G143602, GRMZM2G162670, and GRMZM5G851026) and six hub genes (GRMZM2G162175, GRMZM2G155746, GRMZM2G092325, GRMZM2G328612, AC218148.2_FGT008, and GRMZM5G879127) were identified. Gene expression prediction under density stress was performed using various classifiers including Naïve Bayes, Simple Logistic, KStar, MultiClassClassifier, JRip, LMT, and RandomForest. Utilizing Simple Logistic and LMT models, we achieved an impressive overall accuracy of 100 % in predicting density stress response based on hub gene expression profiles. This highlights the robustness and reliability of our findings, paving the way for developing targeted interventions and breeding strategies to bolster sweet corn's resilience to density stress. Key genes include glycolate oxidase 1, essential for oxidative stress tolerance, and CK2 alpha subunit, involved in signaling pathways for abiotic stress adaptation. Other important proteins, like those from the phosphatidylinositolglycan synthase family, contribute to lipid metabolism and stress signaling. Additionally, uncharacterized genes, LOC103635295 and LOC100274670, are highlighted for their potential roles in stress regulation. The study emphasizes the need for continued research on these genes to enhance crop resilience and productivity.