Computational Methods for Predicting ncRNA-protein Interactions.

Area of Science:

• Molecular Biology
• Bioinformatics

Background:

• RNA-protein interactions (RPIs) are vital for cellular processes.
• Noncoding RNA-protein interactions (ncRPIs) regulate genes and contribute to complex diseases.
• Experimental methods for ncRPIs are costly and time-consuming, necessitating computational prediction.

Purpose of the Study:

• To review recent advancements in computational prediction of ncRPIs.
• To highlight key aspects of dataset construction, feature representation, and machine learning algorithms for ncRPI prediction.
• To identify areas for future improvement in ncRPI prediction methodologies.

Main Methods:

• Dataset construction for ncRPI prediction.
• Sequence and structural feature representation for ncRNAs and proteins.
• Application of machine learning algorithms in predicting ncRPIs.

Main Results:

• Current computational methods have demonstrated success in predicting ncRPIs.
• Existing methods often rely on small benchmark datasets from the PDB.
• There is a need to enhance the generalization performance and robustness of current ncRPI prediction models.

Conclusions:

• Future ncRPI prediction research should focus on effective negative sample set construction.
• Novel and effective feature selection from ncRNA and protein sequences/structures is essential.
• Developing more powerful prediction models is a key future direction for ncRPI research.