AI-Assisted Identification of Candidate Diagnostic Biomarkers for Heart Failure Using Multi-Omics Integration

Ian Pranandi (Universitas Katolik Indonesia Atma Jaya)

Glosains: Jurnal Sains Global Indonesia
Vol. 7 No. 3 (2026): Glosains: Jurnal Sains Global Indonesia

Background: Heart failure (HF) remains a major global health burden associated with high morbidity and mortality. Conventional biomarkers, such as B-type natriuretic peptide (BNP), have limited diagnostic performance because they do not fully capture the molecular complexity of HF. Integrating multi-omics data with artificial intelligence (AI) may improve the identification of novel diagnostic biomarkers. Objective: This study aimed to identify novel diagnostic biomarkers for HF by integrating transcriptomic and proteomic datasets using AI-assisted machine learning. Methods: Public transcriptomic and proteomic datasets from patients with HF and healthy controls were integrated in an in silico bioinformatics study. Differential expression analysis was performed to identify dysregulated genes and proteins, followed by Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment, and protein–protein interaction (PPI) network analyses. Candidate biomarkers were prioritized using Least Absolute Shrinkage and Selection Operator (LASSO), Random Forest, and Extreme Gradient Boosting (XGBoost). Diagnostic performance was evaluated using receiver operating characteristic (ROC) curve analysis. Results: Multi-omics integration identified 1,247 differentially expressed genes, 186 differentially expressed proteins, and 42 overlapping molecules. Enrichment analyses indicated significant involvement in cardiac remodeling, inflammation, oxidative stress, energy metabolism, and calcium signaling. Five hub biomarkers (NPPB, ATP2A2, IL6, PPARGC1A, and TGFB1) were consistently selected by all machine learning models. Individual biomarkers achieved area under the ROC curve (AUC) values ranging from 0.88 to 0.94, whereas the combined five-biomarker panel achieved an AUC of 0.97, with 92.4% sensitivity and 90.1% specificity. Conclusion: AI-assisted multi-omics integration identified five biologically relevant biomarkers with excellent diagnostic performance, providing a promising foundation for future HF diagnostics and precision cardiology.