SeqMG-RPI: A Sequence-Based Framework Integrating Multi-Scale RNA Features and Protein Graphs for RNA-Protein...

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• RNA-protein interactions (RPI) are fundamental to cellular processes.
• Accurate RPI prediction is vital for understanding molecular mechanisms and disease research.
• Existing RPI prediction methods often rely on single features, limiting their performance and generalizability.

Purpose of the Study:

• To develop a novel and accurate sequence-based method for predicting RNA-protein interactions (RPI).
• To improve upon existing RPI prediction methods by integrating diverse feature representations.
• To enhance the understanding of RPIs for advancing molecular and disease research.

Main Methods:

• Proposed SeqMG-RPI, a novel sequence-based RPI prediction method.
• Integrated multi-scale RNA features: multi-channel, k-mer frequency, and k-mer sparse matrix.
• Utilized graph-based protein features for comprehensive protein information capture.
• Developed a novel neural network architecture for integrated feature extraction and RPI prediction.

Main Results:

• SeqMG-RPI demonstrated superior performance compared to existing RPI prediction methods.
• Experimental results across multiple datasets confirmed the method's effectiveness.
• The proposed method exhibited enhanced performance and generalization capabilities.

Conclusions:

• SeqMG-RPI offers a significant advancement in sequence-based RPI prediction.
• The integration of multi-scale RNA and graph-based protein features is effective for RPI prediction.
• The developed neural network architecture provides a robust framework for improving RPI prediction accuracy and generalization.