Inferring protein-protein interactions through high-throughput interaction data from diverse organisms.

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Protein-protein interactions are crucial for cellular functions.
• Protein domains mediate these interactions, making them key targets for computational prediction.
• Evolutionary conservation of protein domains allows cross-species data integration.

Purpose of the Study:

• To develop a computational method for predicting protein-protein interactions using domain-domain interaction probabilities.
• To assess the benefit of integrating data from multiple organisms for improved interaction prediction.

Main Methods:

• A likelihood approach was used to estimate domain-domain interaction probabilities.
• Protein interaction data from Saccharomyces cerevisiae, Caenorhabditis elegans, and Drosophila melanogaster were integrated.
• Predicted protein-protein interactions were evaluated using sensitivity, specificity, Gene Ontology enrichment, and gene expression profiles.

Main Results:

• Integrating domain-domain interaction data from multiple organisms significantly improved protein-protein interaction prediction in S. cerevisiae.
• Cross-species data integration proved more informative than single-organism data for predicting protein interactions.
• The computational program and supplementary materials are publicly available.

Conclusions:

• Predicting protein-protein interactions by integrating domain-domain data across diverse organisms is a powerful approach.
• This method enhances our understanding of cellular processes and protein functions.
• The findings highlight the value of leveraging evolutionary conservation in computational biology.