DeepLocRNA: an interpretable deep learning model for predicting RNA subcellular localization with domain-specific...

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• RNA subcellular localization is crucial for cellular processes and function.
• Trans-acting RNA binding proteins (RBPs) regulate post-transcriptional processes via cis-regulatory RNA motifs.
• Existing prediction methods lack the integration of RBP-binding information.

Purpose of the Study:

• To develop an accurate and interpretable deep-learning model for predicting RNA subcellular localization.
• To incorporate RNA binding protein (RBP) information into RNA localization prediction.
• To provide a tool for understanding RNA localization patterns in different species and RNA types.

Main Methods:

• Developed DeepLocRNA, an interpretable deep-learning model.
• Leveraged a pre-trained multi-task RBP-binding prediction model for fine-tuning.
• Constructed a comprehensive dataset including various RNA types.
• Evaluated model performance on a held-out dataset and performed motif analysis for interpretability.

Main Results:

• Achieved state-of-the-art performance in predicting RNA subcellular localization for mRNA and miRNA.
• Demonstrated strong generalization capabilities across human and mouse RNA.
• Identified key signal factors contributing to predictions through motif analysis.
• The model offers powerful prediction abilities for diverse RNA types and species.

Conclusions:

• DeepLocRNA provides accurate and interpretable RNA subcellular localization predictions.
• The model's integration of RBP-binding information advances the field.
• Offers valuable insights into RNA localization patterns, aiding cellular process understanding.
• A web server is available for public use, facilitating broader research application.