{"database":"GEO","file_versions":[{"headers":{"Content-Type":["application/json"]},"body":{"files":{"Other":["ftp://ftp.ncbi.nlm.nih.gov/geo/series/GSE334nnn/GSE334029/"]},"type":"primary"},"statusCode":"OK","statusCodeValue":200}],"scores":null,"additional":{"omics_type":["Transcriptomics"],"species":["Homo sapiens"],"gds_type":["Expression profiling by high throughput sequencing"],"full_dataset_link":["https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE334029"],"repository":["GEO"],"entry_type":["GSE"],"additional_accession":[]},"is_claimable":false,"name":"EpiFat Coverage: integrative multiomics analysis and machine learning modelling for heart disease","description":"Background: Epicardial adipose tissue (EAT) has emerged as a key factor in the pathogenesis of cardiovascular diseases, including arrhythmias, coronary artery disease (CAD), and heart failure (HF). However, the molecular mechanisms linking EAT characteristics to cardiac pathology remain poorly understood. Objectives: This study aimed to (1) characterize transcriptomic, metabolomic, and lipidomic differences in human hearts across varying levels of epicardial fat coverage (EpiFat Coverage, EFC), and (2) develop machine learning (ML) models to predict cardiovascular risk factors such as CAD, hypertension, and EFC class based on integrated multiomics data. Methods: Thirty-two human donor hearts were analyzed using RNA sequencing, targeted metabolomics, and lipidomics. Donors were classified by BMI, EFC percentage, CAD status, and hypertension history. Differential expression analyses were performed using DESeq2, limma, and Lipidr, while ML classifiers (SVM, random forest, logistic regression, and LDA) and regressors were used to predict EFC, CAD, and hypertension. Results: Distinct molecular signatures were identified for CAD, hypertension, and obesity-related EAT. CAD-associated genes included COL1A1, COL3A1, BGN, and SPP1, while hypertensive samples overexpressed PTGDR and BPIFB4. Overweight and obese EAT samples showed downregulated UCP1 expression. Integrating metabolomic and transcriptomic features improved model accuracy, achieving up to 99% ROC AUC for CAD and 88% for hypertension prediction. EFC classifiers reached 97% accuracy in distinguishing overweight from normal EAT coverage. Conclusions: Integrating multiomics data with ML provides a powerful framework for identifying molecular determinants of EAT-related cardiovascular diseases. The predictive models developed here demonstrate strong potential for precision risk assessment of hypertension, CAD, and EAT burden, paving the way toward personalized cardiac health monitoring.","dates":{"publication":"2026/06/03"},"accession":"GSE334029","cross_references":{"GSM":["GSM9779407"],"GPL":["24676"],"GSE":["334029"],"taxon":["Homo sapiens"]}}