Abstract Body (Do not enter title and authors here): Background:
Dilated cardiomyopathy (DCM) is a major cause of heart failure, characterized by ventricular dilation and reduced systolic function. Autophagy has been increasingly implicated in DCM pathogenesis, but key regulatory genes remain unclear. Leveraging high-throughput sequencing and bioinformatics enables systematic gene screening. This study aimed to identify critical autophagy-related genes in DCM through GEO dataset analysis and machine learning, to provide insights into molecular mechanisms and potential therapeutic targets.
Methods:
We retrieved gene expression data (GSE5406) from the GEO database. After preprocessing with the R package limma, differentially expressed genes (DEGs) were identified between DCM and control samples. By intersecting DEGs with a curated autophagy gene list, we identified 31 autophagy-related DEGs. A heatmap illustrated expression patterns. GO and KEGG enrichment analyses were performed to explore their functions. To identify core feature genes, five models—GLM, RF, SVM, XGB, and LASSO regression—were applied. The overlapping gene across all models, ATF4, was selected. To validate its expression, we used a doxorubicin-induced H9c2 cardiomyocyte model and performed Western blotting.
Results:
Over 1,000 DEGs were detected; 31 were autophagy-related. GO analysis indicated enrichment in unfolded protein binding and cellular responses to stress. KEGG pathways included autophagy, lipid metabolism, and PI3K-Akt signaling. ATF4 emerged as the only consistently identified gene across all five computational models. In vitro validation confirmed ATF4 was significantly downregulated in the DCM model, consistent with in silico findings.
Conclusion:
We identified ATF4 as a potential autophagy-related hub gene in DCM through integrated bioinformatics and machine learning, and validated its downregulation experimentally. ATF4 may modulate autophagy and stress responses, making it a promising diagnostic marker and therapeutic target. These findings provide a foundation for further mechanistic and translational research in DCM.