Inferring circRNA-Disease Associations via Sparse Topological Representation Learning and Dual-View Decoding

Area of Science:

• Biochemistry
• Genomics
• Computational Biology

Background:

• Circular RNAs (circRNAs) are crucial in complex disease development and progression.
• circRNAs show potential as diagnostic and prognostic biomarkers.
• Experimental validation of circRNA-disease associations is costly and time-consuming.

Purpose of the Study:

• To develop an efficient computational method for predicting circRNA-disease associations.
• To overcome limitations of experimental validation for circRNA-disease links.
• To provide a robust tool for uncovering circRNA roles in disease.

Main Methods:

• Constructing fused similarity profiles for circRNAs and diseases using diverse attributes.
• Refining similarity matrices with random walk with restart for local feature capture.
• Employing a sparse-constrained dual-branch graph autoencoder for topological embedding extraction.
• Utilizing an XGBoost classifier to score potential circRNA-disease pairs.

Main Results:

• STRCDA achieved high performance on the CircR2Disease dataset with an AUC of 0.9771 and AUPR of 0.9826.
• Five-fold cross-validation demonstrated the model's robustness.
• 18 out of the top 20 predicted circRNA-disease associations were experimentally validated.

Conclusions:

• STRCDA is an effective computational tool for predicting circRNA-disease associations.
• The method significantly reduces the need for extensive wet-lab validation.
• STRCDA facilitates the discovery of circRNA functions in complex diseases.