GCNCDA

GCNCDA is a computational method to predict potential associations between circular RNAs (circRNAs) and diseases, leveraging the capabilities of deep learning and graph convolutional networks (GCNs). CircRNAs have been recognized for their significant roles in the onset and progression of various diseases, making identifying circRNA-disease associations crucial for understanding complex disease pathogenesis and advancing diagnosis and treatment strategies. However, the intricate mechanisms underlying these associations and the high cost and time demands of biological experiments necessitate efficient computational approaches for discovery.

GCNCDA employs a novel strategy that integrates disease semantic similarity information, disease and circRNA Gaussian Interaction Profile (GIP) kernel similarity information, based on known associations, into a unified descriptor. Utilizing the Fast Learning with Graph Convolutional Networks (FastGCN) algorithm, GCNCDA extracts high-level features from this unified descriptor to accurately predict new circRNA-disease associations using the Forest by Penalizing Attributes (Forest PA) classifier. The method's effectiveness is demonstrated through a 5-fold cross-validation on the circR2Disease benchmark dataset, achieving an accuracy of 91.2%, sensitivity of 92.78%, and an AUC of 90.90%.