MGFmiRNAloc: Predicting miRNA Subcellular Localization Using Molecular Graph Feature and Convolutional Block...

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• MicroRNAs (miRNAs) play crucial roles in various cellular functions.
• Accurate prediction of miRNA subcellular localization is essential but computationally challenging.
• Existing methods for miRNA localization prediction have limitations, necessitating improved approaches.

Purpose of the Study:

• To develop a novel computational method, MGFmiRNAloc, for predicting the subcellular localization of microRNAs (miRNAs).
• To utilize simplified molecular input line entry system (SMILES) format and graphical convolutional networks (GCN) for RNA sequence feature extraction.
• To enhance prediction accuracy through channel attention and spatial attention mechanisms (CBAM).

Main Methods:

• Representing RNA sequences using SMILES format for MGFmiRNAloc.
• Employing GCN to extract atomic nodes and topological structures from RNA sequences, creating molecular map features.
• Integrating CBAM to refine feature extraction for improved information mining.
• Utilizing a prediction module for final subcellular localization detection.

Main Results:

• MGFmiRNAloc demonstrated superior performance compared to existing state-of-the-art methods in predicting miRNA subcellular localization.
• 10-fold cross-validation and independent test set experiments validated the model's effectiveness.
• The proposed atomic-level feature representation successfully addressed limitations of small sample sizes and short miRNA sequences.

Conclusions:

• MGFmiRNAloc offers an accurate and robust method for predicting miRNA subcellular localization.
• The atomic-level feature representation is a promising advancement for analyzing short biological sequences.
• The methodology can be extended to predict the subcellular localization of other types of nucleic acid sequences.