Computational identification of protein complexes from network interactions: Present state, challenges, and the way...

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Protein-protein interactions (PPIs) form macromolecule complexes essential for cellular functions.
• Advances in experimental techniques yield PPI networks, enabling computational analysis and protein complex prediction.
• Understanding protein complexes is crucial for deciphering diverse cellular processes.

Purpose of the Study:

• To systematically review state-of-the-art algorithms for protein complex prediction from PPI networks over the past two decades.
• To categorize existing approaches and compare their advantages and disadvantages.
• To provide a comparative performance analysis of various methods on benchmark datasets.

Main Methods:

• Categorization of algorithms into cluster-quality-based, node affinity-based, and network embedding-based approaches.
• Comparative analysis of eighteen prediction methods using twelve performance measures.
• Evaluation on four widely used benchmark protein-protein interaction networks.

Main Results:

• The study categorizes and contrasts the strengths and weaknesses of different protein complex prediction algorithms.
• A comprehensive performance evaluation of eighteen methods on benchmark PPI networks is presented.
• Limitations of current data and approaches are identified, highlighting areas for future development.

Conclusions:

• The review provides a structured overview of protein complex prediction methodologies.
• Comparative analysis offers insights into the effectiveness of various algorithms.
• Identified limitations and future directions aim to advance computational approaches in this field.