mRSubLoc: A Novel Multi-Label Learning Framework Integrating RNA Large Language Model for mRNA Subcellular...

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• Subcellular localization of messenger RNA (mRNA) is crucial for gene expression regulation and targeted drug development.
• Existing computational models for mRNA localization prediction face challenges in sequence representation and multi-localization tasks.

Purpose of the Study:

• To propose mRSubLoc, a novel multi-label deep learning framework for predicting mRNA subcellular localization.
• To enhance the accuracy and efficiency of mRNA localization prediction, particularly for multi-localization scenarios.

Main Methods:

• Integrated RNA large language model (RNAErnie) with one-hot encoding and Word2Vec embeddings for comprehensive mRNA sequence representation.
• Employed Text Convolutional Neural Network (TextCNN) for local feature extraction and Bidirectional Long Short-Term Memory (BiLSTM) for long-range dependency capture.
• Utilized multi-head self-attention and a multi-layer perceptron (MLP) to fuse features and explore complex dependencies for accurate multi-localization prediction.

Main Results:

• mRSubLoc significantly outperformed state-of-the-art methods on a testing set.
• Achieved high performance across multiple metrics: Aiming (0.7858), Coverage (0.6212), Accuracy (0.6161), Absolute-True (0.3070), and Absolute-False (0.1319).

Conclusions:

• mRSubLoc offers a novel and effective approach for predicting mRNA subcellular localization.
• The framework provides new perspectives for advancing disease diagnosis and drug discovery in biomedical research.