DeepLMI: deep feature mining with a globally enhanced graph convolutional network for robust lncRNA-miRNA interaction...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Long non-coding RNAs (lncRNAs) and microRNAs (miRNAs) are crucial in gene regulation and disease.
• Identifying lncRNA-miRNA interactions is vital for understanding disease mechanisms and therapeutic target discovery.
• Current experimental methods for identifying these interactions are costly and time-consuming, necessitating computational solutions.

Purpose of the Study:

• To develop a novel deep learning framework, DeepLMI, for accurate prediction of lncRNA-miRNA interactions.
• To integrate advanced feature extraction and a globally enhanced graph convolutional network for improved prediction accuracy.
• To provide a reliable computational tool for RNA interaction analysis in disease research.

Main Methods:

• DeepLMI utilizes specialized feature extraction for lncRNAs (sequence pre-training, self-attention) and miRNAs (heterogeneous feature fusion via graph convolutional encoder).
• A Global-Enhanced Graph Convolutional Network (GE-GCN) is employed to model both local and global network topology.
• Learned embeddings for lncRNAs and miRNAs are integrated to predict interaction probabilities.

Main Results:

• DeepLMI demonstrates superior performance compared to existing state-of-the-art methods in predicting lncRNA-miRNA interactions.
• The framework shows consistent accuracy across multiple datasets and evaluation settings.
• DeepLMI exhibits strong robustness, indicating its reliability for RNA interaction analysis.

Conclusions:

• DeepLMI offers a powerful and accurate computational approach for predicting lncRNA-miRNA interactions.
• The framework has significant potential to advance RNA interaction analysis and contribute to disease research.
• The developed method addresses the need for efficient and reliable tools in the field.