Background: Infectious pneumonia remains a major global health challenge with high morbidity and mortality, especially among vulnerable groups. Current diagnostic approaches lack sufficient specificity and accuracy. This study aimed to identify core diagnostic genes, explore their biological functions, and predict potential natural compounds targeting these genes to improve diagnostic and therapeutic strategies. Methods: Gene expression profiles from the GEO database (GSE103119) were analyzed to identify differentially expressed genes (DEGs). Hub genes were selected by integrating protein–protein interaction (PPI) networks and multiple machine learning algorithms. Expression patterns of the identified hub gene were validated in a murine pneumonia model. Reverse network pharmacology was applied to screen natural compounds, followed by molecular docking and molecular dynamics simulations to evaluate binding affinity and complex stability. Results: A total of 2550 DEGs were identified. FCER1A was consistently determined as a hub gene through PPI and machine learning analyses, showing significant downregulation in infectious pneumonia patients. Animal experiments confirmed pronounced reduction of Fcer1a transcription in both lung tissue and whole blood of pneumonia model mice. Two natural compounds, pyrogallol and tectorigenin, were identified as potential ligands for FCER1A. Molecular simulations confirmed stable binding with the target protein, with tectorigenin exhibiting superior binding affinity. Conclusions: This study proposes FCER1A as a promising diagnostic biomarker for infectious pneumonia and suggests tectorigenin as a candidate compound for further therapeutic development.

Chronic kidney disease (CKD) lacks highly specific early diagnostic biomarkers and safe, effective therapeutic options. To address this, we integrated multi-cohort transcriptomics, bioinformatics, and machine learning with reverse network pharmacology, molecular docking, molecular dynamics simulations, and in vivo experiments to identify candidate biomarkers associated with CKD and candidate therapeutic compounds for CKD. Our analyses of the GSE175759 training set and external validation datasets (GSE37171 and GSE66494) using differential expression and WGCNA indicated that CKD is characterized by immune-inflammatory activation and suppressed energy metabolism. Integration of PPI analysis with three machine learning algorithms identified JUN and C3 as candidate diagnostic genes. JUN was downregulated, whereas C3 was upregulated in CKD, both showing promising discriminatory performance in the analyzed datasets. Reverse screening identified Resveratrol and Triptolide as candidate active compounds, and molecular docking together with 100 ns molecular dynamics simulations supported stable binding to JUN/C3 complexes. Given its superior safety profile, Resveratrol was selected for experimental validation. In an adenine-induced CKD rat model, Resveratrol improved renal function, reduced proteinuria, alleviated renal injury and fibrosis, and was associated with reduced activation of MAPK/JNK-c-Jun signaling. In conclusion, this study identifies JUN and C3 as candidate biomarkers associated with CKD and suggests Resveratrol as a promising preclinical intervention candidate targeting the MAPK/JNK-c-Jun axis, providing preliminary computational and preclinical evidence for biomarker discovery and natural medicine-based intervention in CKD.