Assessing and predicting protein interactions using both local and global network topological metrics

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• High-throughput protein interaction data are crucial for biological discovery but suffer from high error rates (false positives/negatives).
• Scalable computational methods are needed to identify and correct errors in these large datasets.

Purpose of the Study:

• To develop a novel computational method for identifying spurious and missing interactions in high-throughput protein interaction data.
• To improve the accuracy and reliability of protein interaction networks for downstream biological analysis.

Main Methods:

• Developed LGTweight, a method integrating local and global topological information of protein pairs.
• Calculated local interacting scores based on common neighbors.
• Computed global interacting scores using globally interacting protein group pairs.
• Combined scores to generate LGTweight for predicting protein interaction possibility.

Main Results:

• Tested on the DIP yeast interaction dataset, LGTweight demonstrated superior performance.
• Interactions ranked highest by LGTweight exhibited greater functional homogeneity and localization coherence.
• Achieved higher sensitivity and precision compared to existing methods via 5-fold cross-validation.

Conclusions:

• LGTweight effectively identifies errors in high-throughput protein interaction data.
• The method enhances the quality of protein interaction networks, supporting more reliable biological insights.
• LGTweight offers improved accuracy and robustness over existing computational approaches.