Machine Learning Models identify Signature Genes as potential Biomarkers for Hypertrophic Cardiomyopathy from Williams Syndrome

Abstract

Background: Hypertrophic cardiomyopathy(HCM) is a complex genetic cardiovascular disease with the pathogenesis is still unclear. Williams syndrome(WS), an autosomal dominant systemic disorder with the phenotype of series congenital heart disease that caused by the missing of low-copy DNA elements. The association between the HCM and WS remains unrevealed. This study aimed to identify the WS-related genes from a special patient by using bioinformatics analysis to clarify insights into the diagnostic value in HCM. Methods and Methods: We collected and summarized the clinical data of a young male patient with WS who initially experienced surgical repairing of coarctation of the aorta(CoA) in his infancy but developed HCM. Whole-Exome Sequencing(WES) data were obtained and intersected with the expression of HCM samples. Bioinformatics analysis, included the consensus cluster analysis, gene set enrichment analysis(GSEA), gene ontology and kyoto encyclopedia of genes and genomes(KEGG) analysis (GO-KEGG) and weight gene correlation analysis(WGCNA) was used to identify the differentially expressed genes(DEGs). Based on machine learning, we selected the characteristic signature genes and miRNA or transcription factor(TF) related networks as potential biomarkers for HCM. Results: A total of 7569 and 3040 DEGs were identified between HCM and WS with control samples and 4 WS related genes(BCL7B, ELN, FZD9, NCF1) were sellected based on our patient. Subsequently, we classified the entire HCM cohort into two subtypes according to the target genes. Moreover,the application of GSEA, GO-KEGG and protein-protein interaction(PPI) network analysis were explored and obtained 22 hub genes with the intersection of clusters and WGCNA. According to the five machine algorithms, 4 features(ST8SIA5, RAP1GAP, PLAU, RGMA) were ascertained as the signatures to construct the HCM diagnostic model. Conclusion: We developed diagnostic signatures to distinguish HCM based on our special patient with WS and the model had certain diagnostic and individual effects. This study sheds light on the potential genetic pathogenesis of HCM and may provide directions for drug screening and personalized therapy in the future.