iGRLCDA: identifying circRNA-disease association based on graph representation learning.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Circular RNAs (circRNAs) are a novel RNA topology with emerging roles in gene regulation.
• circRNAs are implicated in complex diseases, but experimental identification of their associations is time-consuming.
• Understanding circRNA-disease associations is crucial for advancing medical research.

Purpose of the Study:

• To develop a computational method for predicting potential circRNA-disease associations.
• To leverage graph representation learning for accurate circRNA-disease association prediction.
• To reduce the inefficiency and lack of direction in experimental identification of circRNA-disease links.

Main Methods:

• Proposed iGRLCDA, a method combining graph convolution network (GCN) and graph factorization (GF).
• Utilized Gaussian interaction profile (GIP) kernel and disease semantic information to derive hidden features of known circRNA-disease associations.
• Employed a random forest classifier for final prediction of potential circRNA-disease associations.

Main Results:

• Achieved an average area under the receiver operating characteristic curve (AUC) of 0.9289 and an area under the precision-recall curve (AUPRC) of 0.9377 in five-fold cross-validation.
• Demonstrated strong competitiveness compared to existing prediction models on gold standard data.
• 22 out of 30 top predicted circRNA-disease associations were validated by recent literature, indicating high accuracy.

Conclusions:

• iGRLCDA provides a reliable computational approach for predicting circRNA-disease associations.
• The method can significantly reduce the need for extensive and undirected wet-lab experiments.
• iGRLCDA facilitates medical research by offering potential circRNA-disease associations for further investigation.