Base-resolution binding profile prediction of proteins on RNAs with deep learning.

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• RNA-binding proteins are vital for cellular functions and disease pathogenesis.
• Current deep learning methods for protein-RNA interaction prediction using CLIP-seq data do not account for cell-specific gene expression variations.
• This limitation hinders accurate prediction of binding nucleotides and strength across different cell lines.

Purpose of the Study:

• To develop a novel deep learning method, iDeepB, for predicting protein-RNA binding profiles at base resolution.
• To integrate cell-line-specific gene expression profiles into the prediction model.
• To improve the accuracy of protein-RNA interaction prediction across diverse cellular contexts.

Main Methods:

• Constructed expression-aware benchmark datasets using cell-specific RNA-seq and eCLIP-seq data.
• Developed a hybrid deep learning network incorporating multi-head attention.
• Utilized the model for predicting protein binding profiles, analyzing binding motif composition, and quantifying mutation effects.

Main Results:

• iDeepB successfully predicts protein binding profiles on RNAs at base resolution.
• The method demonstrates superior performance compared to existing approaches on newly developed benchmark datasets.
• The model enables analysis of binding motif syntax and functional impact of disease-related mutations.

Conclusions:

• iDeepB offers a significant advancement in predicting protein-RNA interactions by leveraging cell-specific expression data.
• The method enhances the understanding of RNA-associated biological processes and their link to human diseases.
• iDeepB provides a powerful tool for analyzing binding motifs and the functional consequences of genomic variations.