Enhancing LncRNA-miRNA interaction prediction with multimodal contrastive representation learning.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Long non-coding RNAs (lncRNAs) and microRNAs (miRNAs) interactions are vital in complex human diseases.
• Identifying these interactions (LMIs) aids disease diagnosis and treatment.
• Current computational methods struggle with integrating multimodal lncRNA and miRNA data.

Purpose of the Study:

• To propose a novel multimodal contrastive representation learning model (MCRLMI) for accurate LMI prediction.
• To effectively integrate multi-source similarity information and sequence encodings of lncRNAs and miRNAs.
• To improve the prediction performance of LMIs by leveraging multimodal data.

Main Methods:

• Developed a multimodal contrastive representation learning model (MCRLMI).
• Utilized Graph Convolutional Networks (GCN) and Transformers for feature extraction.
• Integrated multichannel attention and contrastive learning for feature fusion.
• Employed a Kolmogorov-Arnold Network (KAN) for final LMI prediction.

Main Results:

• The MCRLMI model demonstrated superior performance compared to existing methods.
• The model successfully integrated structural and semantic information from multimodal data.
• Case studies confirmed the model's potential for identifying novel LMIs.

Conclusions:

• MCRLMI offers a powerful approach for predicting lncRNA-miRNA interactions.
• The model's ability to integrate multimodal data enhances LMI prediction accuracy.
• MCRLMI shows promise for practical applications in disease research.