Increasing confidence of protein interactomes using network topological metrics

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• High-throughput protein-protein interaction (PPI) detection methods yield datasets with significant false positives and negatives.
• Existing experimental approaches are limited in scalability for vast interactomes.
• Complementary computational methods are needed to refine experimental PPI data.

Purpose of the Study:

• To introduce IRAP* (Interaction Repurification via Approximate analysis), a novel computational method.
• To computationally refine experimentally derived protein interactomes by addressing false positives and false negatives.
• To improve the quality and reliability of large-scale PPI datasets.

Main Methods:

• IRAP* employs an iterative computational process to repurify protein interactomes.
• It identifies and removes false positives based on low network topological confidence metrics.
• It identifies and incorporates false negatives using high network topological confidence metrics.

Main Results:

• Application of IRAP* to yeast, fruit fly, and worm interactome datasets.
• Demonstrated reduction in estimated false positive and false negative error rates.
• Computationally refined datasets showed improved functional homogeneity.

Conclusions:

• IRAP* offers a scalable computational solution for enhancing the quality of experimental PPI data.
• The method effectively repurifies large-scale interactomes, reducing errors.
• Confidence indices for PPIs in yeast, fruit fly, and worm are available online.