Machine Learning Advances in Predicting Peptide/Protein-Protein Interactions Based on Sequence Information for Lead...

Area of Science:

• Computational biology
• Drug discovery and development
• Bioinformatics

Background:

• Peptides offer significant advantages in drug discovery.
• High-throughput technologies and artificial intelligence (AI) are enhancing peptide-based drug design.
• Predicting peptide-protein interactions (PepPIs) and protein-protein interactions (PPIs) is crucial for understanding disease mechanisms and identifying lead peptides.

Purpose of the Study:

• To comprehensively review computational models for PepPI and PPI predictions.
• To guide researchers in developing computational frameworks for lead peptide discovery.
• To highlight the integration of various resources for rational drug design.

Main Methods:

• Review of databases for peptide ligands and target proteins.
• Discussion of data formats and feature representations for peptides and proteins.
• Classification and analysis of classical machine learning (ML) and deep learning (DL) methods for prediction models.
• Examination of validation protocols and evaluation metrics for model performance.

Main Results:

• Identification of various computational models for PepPI and PPI prediction.
• Analysis of the strengths and weaknesses of different ML and DL approaches.
• Discussion on the importance of appropriate data representation and validation strategies.
• Overview of available databases and resources.

Conclusions:

• Computational models, particularly ML and DL, are vital tools for PepPI and PPI prediction.
• This review provides a roadmap for researchers to utilize integrated resources for lead peptide discovery.
• Advancements in computational biology accelerate the development of peptide-based therapeutics.