Meta-analysis, WGCNA, and machine learning converge on a four-gene biomarker panel for heat stress tolerance in Solanum lycopersicum.

Heat stress severely compromises growth, reproduction, and yield in Solanum lycopersicum, necessitating the identification of reliable molecular biomarkers for thermotolerance. Here, we performed a comprehensive meta-analysis of transcriptomic datasets encompassing diverse heat-stress experiments and identified 526 differentially expressed genes (DEGs), including 225 upregulated and 301 downregulated candidates. Upregulated genes were strongly enriched in protein folding, unfolded protein binding, and response to heat, whereas downregulated genes were associated with ethylene signaling, secondary metabolism, and growth-related processes. Weighted gene co-expression network analysis (WGCNA) revealed three heat-responsive modules, with the cyan module recapitulating the canonical heat-shock response and negatively correlated modules highlighting coordinated growth repression. By integrating meta-DEGs and WGCNA hubs with two complementary machine-learning methods - Support Vector Machine-Recursive Feature Elimination (SVM-RFE) and Least Absolute Shrinkage and Selection Operator (LASSO) identified a four-gene signature: Solyc09g074500, Solyc01g102960 (class IV sHSP), Solyc07g042230 (ERF-H9), and Solyc02g091990 (ACS3). This panel achieved 98.5% classification accuracy and individual AUC values > 0.95, demonstrating high classification performance in the analyzed datasets. The signature is associated with coordinated chaperone induction alongside reduced expression of ethylene biosynthesis and growth-related pathways under heat stress. These four genes provide promising candidates for further functional validation and future incorporation into breeding or genome-editing strategies.