A supervised protein complex prediction method with network representation learning and gene ontology knowledge.

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Protein complexes are crucial for understanding cellular organization and function.
• Predicting protein complexes from protein-protein interaction (PPI) networks is a key research area.
• Effectively utilizing known protein complex information remains a challenge in prediction methods.

Purpose of the Study:

• To develop a supervised learning method that fully leverages known protein complex information for accurate prediction.
• To address noise and improve the reliability of protein complex identification from PPI networks.

Main Methods:

• Constructed a weighted PPI network integrating gene ontology knowledge and topological information.
• Extracted topological features of known complexes to train a supervised model (SVCC).
• Employed network representation learning and a random forest model for candidate complex classification.

Main Results:

• The proposed method effectively predicts new protein complexes using known complex information.
• Evaluated performance on two public PPI datasets, demonstrating improved prediction accuracy.
• The method successfully reduces noise in PPI networks for better complex identification.

Conclusions:

• The developed method significantly enhances protein complex recognition compared to existing approaches.
• Predicted protein complexes exhibit high biological significance, validated through analysis.
• This approach offers a robust framework for computational protein complex prediction.