DAmiRLocGNet: miRNA subcellular localization prediction by combining miRNA-disease associations and graph...

Area of Science:

• Biochemistry and Molecular Biology
• Bioinformatics
• Genomics

Background:

• MicroRNAs (miRNAs) are key post-transcriptional regulators influencing numerous physiological processes.
• Understanding miRNA subcellular localization is critical for elucidating their biological functions.
• Existing computational methods struggle with comprehensive miRNA functional representation.

Purpose of the Study:

• To develop a novel computational model for accurate prediction of miRNA subcellular localization.
• To address limitations in existing methods regarding miRNA functional representation.
• To leverage miRNA-disease associations and semantic information for improved localization prediction.

Main Methods:

• A Graph Convolutional Network (GCN) and Autoencoder (AE) based model, DAmiRLocGNet, was developed.
• Features were constructed using miRNA sequence, miRNA-disease associations, and disease semantic information.
• GCN captured network structure information; AE captured sequence semantics.

Main Results:

• DAmiRLocGNet demonstrated superior performance compared to existing computational approaches.
• The model effectively extracts implicit features through GCNs.
• Accurate prediction of miRNA subcellular localization was achieved.

Conclusions:

• DAmiRLocGNet offers a powerful tool for identifying miRNA subcellular localization.
• The model has potential applications for other non-coding RNAs.
• It facilitates deeper investigation into the functional mechanisms of miRNA localization.