NARVGA: A Hybrid Framework Integrating Matrix Factorisation and Adversarial Graph Learning for circRNA-Disease Association Prediction

Abstract

Simple Summary Circular ribonucleic acid molecules help regulate gene activity and may contribute to many diseases. However, laboratory experiments can examine only a small fraction of the possible links between these molecules and diseases, making it difficult to identify the most promising candidates for further study. This study developed a computer-based method that combines several types of biological information to predict likely relationships between circular ribonucleic acids and diseases. When tested on a widely used collection of known associations, the method showed a strong ability to distinguish known associations from unconfirmed ones. It also performed well on two related collections involving other types of ribonucleic acid. In a liver cancer case study, 19 of the 20 highest-ranked circular ribonucleic acids were supported by published studies, while one remained a potentially new candidate. These findings suggest that the method can help researchers select promising associations for laboratory testing, reduce unnecessary experimental screening, and support the discovery of disease-related markers and treatment targets.