Publication

Abstract : Biomarker identification plays a crucial role in understanding disease mechanisms and advancing precision medicine. Existing computational models for biomarker-disease associations predominantly focus on either lncRNA-disease or miRNA-disease relationships, often lacking interpretability and comprehensive network insights. In this study, we propose BioLinker, a hybrid graph-based model that integrates lncRNA, miRNA, and disease associations to enhance biomarker discovery. Our approach leverages graph neural networks (GNNs) and an attention-based ranking mechanism to identify significant biomarker-disease relationships with high interpretability. By mapping numerical identifiers to human-readable biomarker names, our model provides biologically meaningful insights, aiding researchers and clinicians in disease diagnosis and therapeutic target discovery. Experimental validation demonstrates the reliability of BioLinker, achieving an AUC of 99%, ensuring robustness in predictive performance. The proposed model has the potential to facilitate early disease detection, biomarker ranking, and drug discovery, contributing to the advancement of bioinformatics and computational medicine