DCDA: CircRNA-Disease Association Prediction with Feed-Forward Neural Network and Deep Autoencoder

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Circular RNAs (circRNAs) are crucial in biological processes and implicated in human diseases.
• Identifying circRNA-disease associations is vital for early diagnosis and therapeutic development.
• Conventional experimental methods for detecting these associations are time-consuming and expensive.

Purpose of the Study:

• To develop an efficient computational method for predicting circRNA-disease associations.
• To identify potential circRNA biomarkers and drug targets for various diseases.
• To overcome the limitations of traditional experimental approaches.

Main Methods:

• A deep learning-based predictor, DCDA, was developed.
• Multiple data sources were integrated to extract circRNA and disease features.
• A deep autoencoder was employed to reveal hidden feature codings.
• A deep neural network was used to predict the association score.

Main Results:

• The DCDA model achieved a high prediction performance.
• Fivefold cross-validation demonstrated the model's effectiveness.
• The model outperformed existing state-of-the-art prediction methods.
• An AUC score of 0.9794 was obtained on the benchmark dataset.

Conclusions:

• DCDA offers a powerful and accurate computational approach for predicting circRNA-disease associations.
• The methodology facilitates the discovery of novel circRNA-disease relationships.
• This work contributes to advancing the use of circRNAs in clinical diagnostics and therapeutics.