Decoding the cytokine code for heart failure based on bioinformatics, machine learning and Bayesian networks

Abstract

Background

Despite maximal pharmacological treatment guided by clinical guidelines, the prognosis of heart failure (HF) remains poor, posing a significant public health burden. This necessitates uncovering novel pathological and cardioprotective pathways. Targeting cytokines presents a promising therapeutic strategy for HF, yet their intricate mechanisms in HF progression remain obscure.

Methods

HF datasets were obtained from the GEO database. Cytokine-related genes were identified through WGCNA and the CytReg database. GO and KEGG enrichment analyses were conducted using the clusterProfiler package. Reactome pathway enrichment analysis and Bayesian regulatory network construction were performed using the CBNplot package. Key genes were identified via LASSO regression and RF algorithms, with diagnostic accuracy evaluated by ROC curves. Potential therapeutic drugs were predicted using the DSigDB database, and immune cell infiltration was assessed with the CIBERSORT package.

Results

We identified 13 cytokine-related genes associated with HF. Enrichment analyses indicated these genes mediate inflammatory responses and immune cell recruitment. Bayesian network analysis revealed two cytokine regulatory chains: IL34-CCL5-CCL4 and IL34-CCL5-CXCL12. Machine learning algorithms identified five key cytokine genes: CCL4, CCL5, CXCL12, CXCL14, and IL34. The DSigDB database predicted 47 potential therapeutic drugs, including Proscillaridin. Immune infiltration analysis showed significant differences in seven immune cell types between HF and healthy samples.

Conclusion

Our study provides insights into cytokines' molecular mechanisms in HF pathophysiology and highlights potential immunomodulatory strategies, gene therapies, and candidate drugs. Future research should validate these findings in clinical settings to develop effective HF therapies.