Recent advances in deep learning for protein-protein interaction: a review

Area of Science:

• Computational Biology
• Artificial Intelligence
• Bioinformatics

Background:

• Protein-protein interactions (PPIs) are crucial for biological functions.
• Deep learning (DL) is revolutionizing computational biology and PPI research.
• A comprehensive review of recent DL advancements in PPI prediction is needed.

Purpose of the Study:

• To assess deep learning progress in protein-protein interaction (PPI) prediction from 2021 to 2025.
• To evaluate core DL architectures and pioneering approaches in PPI analysis.
• To identify challenges and future perspectives in the field.

Main Methods:

• Review of deep learning architectures including GNNs, CNNs, and RNNs.
• Evaluation of advanced methods: Transformers, multi-task learning, multimodal integration, and transfer learning (BERT, ESM).
• Analysis of algorithms for data imbalance, variations, and feature sparsity.

Main Results:

• Detailed assessment of deep learning models for PPI prediction (2021-2025).
• Examination of methods addressing data challenges and industry-specific issues.
• Identification of emerging trends and future research avenues.

Conclusions:

• Deep learning shows significant progress in PPI prediction, enhancing analytical methods.
• Addressing data challenges and exploring novel applications are key for future development.
• This review provides a valuable reference for computational biology and deep learning researchers.