BCMCMI: A Fusion Model for Predicting circRNA-miRNA Interactions Combining Semantic and Meta-path.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Circular RNAs (circRNAs) and microRNAs (miRNAs) are key players in disease pathogenesis.
• Understanding circRNA-miRNA interactions (CMI) is vital for disease diagnosis and treatment.
• Traditional experimental methods for CMI prediction are resource-intensive and time-consuming.

Purpose of the Study:

• To develop an accurate and efficient computational model for predicting circRNA-miRNA interactions.
• To leverage sequence and network information for enhanced CMI prediction.
• To provide a valuable computational tool complementing experimental approaches in disease research.

Main Methods:

• Utilized the BERT algorithm for extracting semantic sequence features from circRNAs and miRNAs.
• Constructed a heterogeneous network incorporating cosine similarity and known CMI data.
• Employed Metapath2vec for capturing topological and similarity features via network random walks.
• Applied the XGBoost classifier for the final prediction of potential CMIs.

Main Results:

• The BCMCMI model demonstrated superior performance in CMI prediction compared to existing state-of-the-art methods.
• Evaluated on two benchmark datasets, BCMCMI achieved significant accuracy.
• t-SNE visualization confirmed the effectiveness of feature extraction and distribution.

Conclusions:

• BCMCMI offers a highly accurate computational approach for predicting circRNA-miRNA interactions.
• The model effectively integrates sequence and network-based features for robust prediction.
• BCMCMI serves as an efficient and valuable complement to traditional experimental methods in CMI research.