HGECDA: A Heterogeneous Graph Embedding Model for CircRNA-Disease Association Prediction

Area of Science:

• Biochemistry
• Genomics
• Computational Biology

Background:

• Circular RNAs (circRNAs) exhibit specific expression patterns in disease tissues, making them potential diagnostic biomarkers.
• Understanding circRNA-disease associations is crucial for elucidating disease mechanisms and identifying therapeutic targets.
• Current prediction methods often overlook the complex interplay involving microRNAs (miRNAs).

Purpose of the Study:

• To develop a novel computational method, HGECDA, for predicting circRNA-disease associations.
• To incorporate heterogeneous biological information, including miRNA associations, into the prediction model.
• To enhance the accuracy and utility of circRNA-disease association prediction.

Main Methods:

• Construction of a heterogeneous graph network integrating circRNA-miRNA-disease associations.
• Utilizing meta-path-based random walks for sampling heterogeneous graph information.
• Employing a path embedding model with skip-gram and negative sampling for initial feature vector generation.
• Designing the CosMulformer model with linearized self-attention and Hadamard product for final prediction.

Main Results:

• MicroRNA data significantly enriches the feature space for circRNA-disease association prediction.
• The CosMulformer model effectively captures deep local interaction features.
• HGECDA outperforms seven existing state-of-the-art methods on benchmark datasets.
• Case studies on breast and colorectal cancers validate HGECDA's practical applicability.

Conclusions:

• HGECDA provides a robust framework for predicting circRNA-disease associations by leveraging miRNA data.
• The integration of heterogeneous biological information is vital for accurate prediction.
• This method holds promise for advancing our understanding of disease pathogenesis and biomarker discovery.