LCKGCN: Identifying Potential Circrna–Disease Associations Based on Large Convolutional Kernel and Graph Convolutional Network

circRNA plays an important role in multicellular organisms, and circRNA has important functions in many complex diseases. CircRNA-disease association prediction helps to find potential biomarkers of diseases. However, existing methods cannot effectively use information from multi-source data and cannot effectively extract important features of signatures. In this paper, we propose a circRNA-disease association prediction method LCKGCN (prediction circRNA–disease associations based on Large Convolutional Kernel and Graph Convolutional Network), which uses large convolution kernels to effectively redefine the information distribution in multi-source data, and uses graph convolution networks to effectively obtain potentially important features in multi-source data. The results of five-fold and ten-fold cross-validation in the data show that LCKGCN is better than existing methods. We used LCKGCN to predict 10 disease-associated circRNAs, 8 of which have been reported in relevant literature, proving that our method can effectively predict disease-associated circRNAs.