Method for Essential Protein Prediction Based on a Novel Weighted Protein-Domain Interaction Network

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Protein-protein interaction (PPI) networks are crucial for understanding biological processes, but current models for identifying key proteins face challenges due to network inaccuracies.
• Existing computational models struggle with false positives, false negatives, and incomplete data in PPI networks, limiting their predictive accuracy for key protein identification.

Purpose of the Study:

• To develop a novel prediction model, WPDINM, for accurate key protein detection.
• To leverage a weighted protein-domain interaction (PDI) network by integrating multiple data sources for enhanced prediction.

Main Methods:

• Constructed a weighted PPI network by combining gene expression data and network topology.
• Developed a weighted domain-domain interaction (DDI) network from the PDI network.
• Integrated weighted PPI, DDI, and PDI networks to create a comprehensive weighted PDI network.
• Applied a PageRank-based iterative algorithm incorporating topological and biological features (subcellular localization, orthology) to estimate protein criticality.

Main Results:

• WPDINM achieved high prediction accuracy rates: 90.19% (top 1%), 81.96% (top 5%), 70.72% (top 10%), 62.04% (top 15%), 55.83% (top 20%), and 51.13% (top 25%).
• The model significantly outperformed 12 traditional state-of-the-art competing measures in key protein identification.
• Experimental validation confirmed the superior predictive performance of WPDINM.

Conclusions:

• WPDINM demonstrates a robust and accurate approach for identifying key proteins.
• The proposed weighted PDI network and PageRank-based algorithm offer a significant advancement in computational biology.
• This method holds potential for future developments in key protein identification and related biological research.