Predicting circRNA-Disease Associations through Multisource Domain-Aware Embeddings and Feature Projection Networks.

Area of Science:

• Biochemistry
• Genomics
• Computational Biology

Background:

• Circular RNAs (circRNAs) play a significant role in various diseases.
• Accurate prediction of circRNA-disease associations is vital for understanding disease mechanisms.
• Existing methods face challenges due to limited known associations and high experimental costs.

Purpose of the Study:

• To develop a computational method (MNDCDA) for predicting circRNA-disease associations.
• To integrate multiple biological data sources for enhanced prediction accuracy.
• To provide a cost-effective tool for identifying novel circRNA-disease links.

Main Methods:

• Constructed four similarity networks using comprehensive biometric data.
• Employed a neighborhood-aware embedding model to capture structural information.
• Utilized deep feature projection networks for high-order feature interaction learning.
• Applied a bilinear decoder to identify novel circRNA-disease associations.

Main Results:

• The MNDCDA model achieved an Area Under the Curve (AUC) of 0.9070 on a benchmark dataset.
• Case studies validated 25 out of 30 predicted circRNA-disease pairs through experiments and literature.
• Demonstrated robustness and effectiveness in predicting circRNA-disease associations.

Conclusions:

• MNDCDA is a robust computational tool for predicting circRNA-disease associations.
• The method offers valuable insights into disease mechanisms.
• It effectively reduces the cost and effort of biological experiments.