Graph embedding ensemble methods based on the heterogeneous network for lncRNA-miRNA interaction prediction.

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• Long non-coding RNAs (lncRNAs) and microRNAs (miRNAs) are key regulators of gene expression.
• Identifying lncRNA-miRNA interactions (LMIs) is crucial for understanding biological processes and diseases.
• Traditional experimental methods for LMI identification are costly and time-consuming, necessitating computational approaches.

Purpose of the Study:

• To develop novel computational methods for predicting lncRNA-miRNA interactions (LMIs).
• To leverage graph embedding and ensemble learning to exploit network structure and sequence information.
• To improve the accuracy and efficiency of LMI prediction compared to existing methods.

Main Methods:

• Constructed a heterogeneous network integrating lncRNA-lncRNA and miRNA-miRNA sequence similarities with known LMIs.
• Applied graph embedding techniques to learn feature representations of lncRNAs and miRNAs.
• Developed two ensemble strategies: GEEL-PI (integrating base predictors) and GEEL-FI (using a deep attention neural network).

Main Results:

• Both GEEL-PI and GEEL-FI demonstrated superior performance over state-of-the-art methods in LMI prediction.
• Experimental validation confirmed the effectiveness of the proposed ensemble strategies.
• Case studies indicated the capability of GEEL-PI and GEEL-FI in identifying novel lncRNA-miRNA associations.

Conclusions:

• Graph embedding and ensemble learning effectively integrate heterogeneous information for LMI prediction.
• GEEL-PI and GEEL-FI offer efficient and accurate solutions for predicting lncRNA-miRNA interactions.
• These methods hold promise for advancing research in gene regulation and disease mechanisms.