Machine learning combined with multi-omics analysis: identifying nucleotide metabolism-associated immune genes and validating their functions in cardiomyopathy

Cardiomyopathy (CMP) is a heterogeneous group of myocardial disorders with diverse etiologies, posing a significant threat to patient health and quality of life. Accumulating studies have emphasized the role of nucleotide metabolism in CMP pathogenesis, such as regulating myocardial energy homeostasis and inflammatory responses. However, the association between nucleotide metabolism-associated genes (NMGs) and immune dysregulation in CMP remains unclear, and the diagnostic and therapeutic potential of nucleotide metabolism-associated genes (NMGs) in Cardiomyopathy (CMP) has not been fully explored. This study was designed to detect NMGs linked to CMP, dissect their roles in disease progression (especially immune regulation), and uncover novel diagnostic biomarkers and therapeutic targets. We collected RNA sequencing data of CMP from the Gene Expression Omnibus (GEO). Using R, differential expression analysis and weighted gene co-expression network analysis(WGCNA) were carried out, and the resulting data were cross-referenced with a nucleotide metabolism gene set. We employed functional enrichment analysis and the connectivity map (CMap) to identify both differentially expressed genes (DEGs) and potential therapeutic agents.Key immune-associated genes were filtered out using LASSO regression, SVM-RFE, and random forest algorithms. CIBERSORT was utilized to analyze the infiltration patterns and correlation of immune cells, while we conducted in vivo experiments with the doxorubicin-induced cardiomyopathy mouse model (DCMM) to validate the core genes. Additionally, we conducted molecular docking to assess the binding affinity existing between core genes and candidate compounds. In silico ADMET analysis was subsequently performed to evaluate the pharmacokinetic properties, druggability, and safety profiles of the candidate compounds. Thirty-six candidate genes were identified, with ASPN, LUM, and HTRA1 emerging as pivotal immune-associated genes. These three genes exhibited strong correlations with immune cell infiltration and showed favorable diagnostic utility for CMP: the diagnostic the area under curve (AUC) of the three genes was 0.952 in the training cohort and 0.935 in external validation set. Furthermore, based on the CMap database, LY-2,183,240 and rigosertib were predicted as potential therapeutic drugs, and molecular docking verified their robust binding affinity with the core genes. In silico ADMET analysis further demonstrated that both LY-2,183,240 and rigosertib possessed favorable pharmacokinetic and safety profiles, including satisfactory gastrointestinal absorption, absence of P-glycoprotein substrate characteristics, and no hERG channel blocking activity, supporting their potential clinical applicability. Additionally, we conducted in vivo experiments with the DCMM, which confirmed these key genes were significantly dysregulated in CMP. This study underscores the function of NMGs in CMP, while pinpointing ASPN, LUM, and HTRA1 as potential diagnostic biomarkers and therapeutic targets. Lower norm_cs scores indicate that LY-2,183,240 and rigosertib have promising therapeutic potential, and molecular docking confirmed stable binding between these two compounds and the three hub proteins. In addition, ADMET property analysis demonstrates that both LY-2,183,240 and rigosertib possess favorable pharmacokinetic and safety profiles, including good gastrointestinal absorption, no P-glycoprotein substrate characteristics, and no hERG channel blockade, which further supports their potential for clinical application in CMP targeted therapy. These findings provide a basis for targeted therapy of CMP, although further experimental verification is required.