Integrative meta-transcriptomics and machine learning reveal defence networks and regulatory hubs in response to tomato bacterial pathogens

Tomato ( Solanum lycopersicum ) is an economically important vegetable crop, however, its production is severely affected by bacterial pathogens, including Clavibacter michiganensis subsp. michiganensis ( Cmm ) and Pseudomonas syringae . Plant defence responses are regulated by complex hormonal networks, particularly salicylic acid and jasmonic acid, which activate partially overlapping and context-dependent signalling pathways. Although numerous transcriptomic studies have investigated hormone- and pathogen-associated responses, conserved transcriptional signatures underlying tomato defence remain incompletely understood. To address this, we performed an integrative meta-analysis of massive RNA-seq data using five complementary approaches (fixed-effects, random-effects, Rank Product, Stouffer’s Z-score, and Fisher’s method), identifying 178 upregulated and 177 downregulated consensus DEGs representing a core hormone-responsive signature. Functional enrichment analysis revealed extensive metabolic and regulatory reprogramming involving amino acid metabolism, lipid-associated processes, vesicle trafficking, redox regulation and defence-associated signalling pathways. Protein-protein interaction analysis highlighted highly connected genes, including P5CS ( Solyc08G000796 ) and glutathione reductase ( Solyc09G002031 ), which were associated with proline metabolism and redox homeostasis. Weighted gene co-expression network analysis (WGCNA) combined with machine learning-based feature selection enabled the prioritization of integrative candidate genes, including a xyloglucan endotransglucosylase/hydrolase ( Solyc09G000233 ) and other poorly characterized candidates. In addition, the Random Forest model independently prioritized several highly important features, including Solyc10G001545 , Solyc10G002708 , Solyc07G002639 , Solyc12G001296 which showed strong associations with defence-related transcriptional variation. Collectively, our findings suggest that tomato defence responses involve a complex regulatory landscape integrating metabolic adjustment, signaling processes, redox balance, and cell wall-associated responses. The combination of transcriptomic meta-analysis, network biology, and machine learning provides a framework for prioritizing candidate genes and generating hypotheses for future functional studies aimed at understanding tomato defence responses.