RMDNet: RNA-aware dung beetle optimization-based multi-branch integration network for RNA-protein binding sites...

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• RNA-binding proteins (RBPs) are vital for gene regulation, and their dysregulation is implicated in diseases like cancer and neurodegeneration.
• Experimental methods for identifying RNA-protein binding sites (e.g., CLIP-seq) are effective but resource-intensive.
• There is a need for efficient computational tools to predict RNA-protein interactions.

Purpose of the Study:

• To develop a deep learning framework, RMDNet, for accurate prediction of RNA-protein binding sites.
• To integrate multi-scale sequence features and RNA structural information for improved prediction.
• To enhance model interpretability and assess its utility in disease-related research.

Main Methods:

• RMDNet employs a hybrid architecture combining Convolutional Neural Networks (CNNs), CNN-Transformer, and Residual Networks (ResNets).
• RNA secondary structure graphs are processed using a graph neural network with DiffPool for structural feature extraction.
• An improved dung beetle optimization algorithm adaptively weights feature fusion during inference.

Main Results:

• RMDNet significantly outperformed existing state-of-the-art models on the RBP-24 benchmark dataset.
• The model demonstrated strong generalization capabilities on the RBP-31 dataset.
• Ablation studies confirmed the effectiveness of individual RMDNet modules, and biological interpretability was validated through motif extraction and a case study on YTHDF1.

Conclusions:

• RMDNet provides a robust and interpretable computational approach for predicting RNA-protein binding sites.
• The framework holds significant potential for advancing research into disease mechanisms and identifying therapeutic targets.
• The source code is publicly available for further research and application.