This study presents an updated web server, CircDA, which is a computational model for predicting circRNA-associated diseases using a graph Markov neural network (GMNN) algorithm. CircDA integrates multisource biological data, including omics data, to construct a heterogeneous biomolecular association network. The model uses matrix factorization and a convolutional network to learn deep feature representations and combines a graph autoencoder with variational reasoning to alternately learn features and propagate labels. Case studies on human hepatocellular carcinoma (HCC) tissue data demonstrate that CircDA can identify missing associations between known circRNAs and diseases. RT-qPCR experiments on HCC tissue samples validated the prediction results, showing that five out of ten predicted circRNAs were significantly differentially expressed. The web server provides an intuitive interface for users to input circRNA sequences and predict associated diseases, along with downloadable trained models and Python code. The study highlights the effectiveness of CircDA in predicting circRNA-disease associations and its potential for guiding biological experiments and disease research.This study presents an updated web server, CircDA, which is a computational model for predicting circRNA-associated diseases using a graph Markov neural network (GMNN) algorithm. CircDA integrates multisource biological data, including omics data, to construct a heterogeneous biomolecular association network. The model uses matrix factorization and a convolutional network to learn deep feature representations and combines a graph autoencoder with variational reasoning to alternately learn features and propagate labels. Case studies on human hepatocellular carcinoma (HCC) tissue data demonstrate that CircDA can identify missing associations between known circRNAs and diseases. RT-qPCR experiments on HCC tissue samples validated the prediction results, showing that five out of ten predicted circRNAs were significantly differentially expressed. The web server provides an intuitive interface for users to input circRNA sequences and predict associated diseases, along with downloadable trained models and Python code. The study highlights the effectiveness of CircDA in predicting circRNA-disease associations and its potential for guiding biological experiments and disease research.