An ensemble approach for CircRNA-disease association prediction based on autoencoder and deep neural network

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Circular RNAs (circRNAs) are crucial regulators of biological processes.
• circRNA-disease associations are implicated in numerous chronic human diseases.
• Experimental identification of these associations is challenging and time-consuming.

Purpose of the Study:

• To develop an efficient computational method for predicting circRNA-disease associations.
• To leverage machine learning for understanding disease pathogenesis through circRNA interactions.

Main Methods:

• An ensemble approach (AE-DNN) combining autoencoders and deep neural networks was developed.
• Features were constructed using circRNA sequence similarity, disease semantic similarity, and Gaussian interaction profile kernel similarities.
• A deep autoencoder extracted high-level features, which were then used by a deep neural network for prediction.

Main Results:

• The AE-DNN model achieved high performance in cross-validation experiments.
• Achieved Area Under the Curve (AUC) scores of 0.9392 (5-fold) and 0.9431 (10-fold).
• Case studies demonstrated the model's robustness and predictive power.

Conclusions:

• AE-DNN is a robust and effective computational tool for predicting circRNA-disease associations.
• This approach can significantly aid in diagnosing disease pathogenesis.
• The model offers a valuable alternative to experimental methods for identifying circRNA-disease links.