Partial correlation coefficient between distance matrices as a new indicator of protein-protein interactions

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Protein-protein interactions (PPIs) are crucial for cellular functions.
• Computational prediction of PPIs is a significant challenge in bioinformatics.
• Co-evolution-based methods are emerging as promising approaches for PPI prediction.

Purpose of the Study:

• To introduce a novel measure, the partial correlation coefficient, for quantifying protein co-evolution.
• To propose and evaluate the use of this coefficient for predicting protein-protein interactions.

Main Methods:

• Developed a partial correlation coefficient to assess the degree of co-evolution between protein sequences.
• Compared the predictive accuracy of the partial correlation coefficient with existing methods like mirror tree and projection methods.
• Utilized an R script for implementing the prediction method.

Main Results:

• The partial correlation coefficient demonstrated effectiveness in reducing false positives in protein-protein interaction prediction.
• An increase in false negatives was observed when using the partial correlation coefficient compared to other methods.
• The proposed method offers an alternative for refining PPI predictions by minimizing erroneous positive interactions.

Conclusions:

• The partial correlation coefficient is a valuable new metric for analyzing protein co-evolution.
• This method shows potential for improving the accuracy of computational protein-protein interaction prediction, particularly in reducing false positives.
• Further research may explore optimizing the method to balance false positive and false negative rates.