PMiSLocMF: predicting miRNA subcellular localizations by incorporating multi-source features of miRNAs.

Area of Science:

• Bioinformatics
• Molecular Biology
• Computational Biology

Background:

• MicroRNAs (miRNAs) are key regulators in biological processes.
• Determining miRNA subcellular localization is vital for understanding their functions.
• Traditional methods are costly, necessitating efficient computational approaches.

Purpose of the Study:

• To develop a novel computational method, PMiSLocMF, for predicting miRNA subcellular localization.
• To address the challenge of multiple subcellular localizations for individual miRNAs using a multi-label classification approach.

Main Methods:

• PMiSLocMF integrates diverse miRNA properties: sequence, functional similarity, and associations with diseases, drugs, and mRNAs.
• Feature extraction employs node2vec and graph attention auto-encoder (GATE) algorithms, generating five distinct feature types.
• A self-attention mechanism and fully connected layers are utilized for prediction.

Main Results:

• PMiSLocMF achieved high performance with accuracy >0.83, average AUC >0.90, and AUPR >0.77.
• The method surpassed all previously reported computational approaches on the same dataset.
• Utilizing all feature types and incorporating GATE and self-attention layers significantly enhanced prediction performance.

Conclusions:

• PMiSLocMF offers a powerful and accurate computational tool for predicting miRNA subcellular localization.
• The study highlights the importance of integrating multiple data sources and advanced deep learning techniques for improved miRNA analysis.
• The findings contribute to a deeper understanding of miRNA roles in biological systems.