Deep Matrix Factorization Improves Prediction of Human CircRNA-Disease Associations

Area of Science:

• Biochemistry
• Genomics
• Computational Biology

Background:

• Circular RNAs (circRNAs) are increasingly recognized for their roles in biological processes.
• Dysregulation and mutations in circRNAs are linked to various diseases, highlighting their diagnostic potential.
• Accurate prediction of circRNA-disease associations is vital for disease diagnosis and management.

Purpose of the Study:

• To develop a novel computational method, DMFCDA (Deep Matrix Factorization CircRNA-Disease Association), for inferring potential circRNA-disease associations.
• To address the challenges in representing complex data structures and underlying features for circRNA-disease association prediction.
• To leverage both explicit and implicit feedback for more accurate predictions.

Main Methods:

• DMFCDA utilizes deep matrix factorization to learn latent representations of circRNAs and diseases.
• The method incorporates a projection layer for automatic feature learning.
• Multi-layer neural networks are employed to model non-linear associations within the data.

Main Results:

• DMFCDA demonstrated efficient inference of circRNA-disease associations on two datasets.
• Performance was validated using leave-one-out and 5-fold cross-validation, showing high AUC values.
• Case studies confirmed the accuracy and reliability of DMFCDA's predictions.

Conclusions:

• DMFCDA provides an accurate and efficient computational approach for predicting circRNA-disease associations.
• The method's ability to model complex, non-linear relationships enhances its predictive power.
• DMFCDA holds significant potential for advancing circRNA-based diagnostics and understanding disease mechanisms.