Protein Function Prediction Based on PPI Networks: Network Reconstruction vs Edge Enrichment.

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• High-throughput technologies generate vast amounts of protein-protein interaction (PPI) data.
• Data quality issues (noise, incompleteness) in PPI networks hinder accurate protein function prediction.
• Network reconstruction and edge enrichment are strategies to improve PPI network quality.

Purpose of the Study:

• To comparatively evaluate the performance of network reconstruction and edge enrichment strategies for PPI networks.
• To assess the impact of different protein similarity metrics (local, global, sequence) on these strategies.
• To determine the optimal pre-processing scheme for enhancing protein function prediction.

Main Methods:

• Applied three protein similarity metrics (local, global, sequence) for network reconstruction and edge enrichment.
• Utilized two real-world PPI datasets for evaluation.
• Compared the performance of network reconstruction, edge enrichment, and original PPI networks.

Main Results:

• Edge enrichment significantly outperformed both network reconstruction and original PPI networks.
• Among similarity metrics, sequence similarity demonstrated superior performance for edge enrichment.
• The findings highlight the effectiveness of edge enrichment over network reconstruction.

Conclusions:

• Edge enrichment is a more effective pre-processing strategy than network reconstruction for PPI networks.
• Sequence similarity is the preferred metric for edge enrichment to improve protein function prediction.
• This study provides guidance for biologists in selecting optimal pre-processing methods for PPI data analysis.