Weighted mutual information analysis substantially improves domain-based functional network models.

Area of Science:

• Bioinformatics
• Systems Biology
• Computational Biology

Background:

• Functional protein-protein interaction (PPI) networks are crucial for understanding molecular pathways in complex phenotypes and human diseases.
• Inferring PPI networks often relies on extrapolating domain-domain interactions (DDIs) from known PPIs, treating protein domains as functional units.

Purpose of the Study:

• To present a novel method for inferring accurate functional PPIs by analyzing the similarity of protein domain composition.
• To evaluate the performance of this method against existing domain-based network inference approaches.

Main Methods:

• Utilized weighted mutual information (MI) to quantify domain composition similarity between proteins.
• Assigned differential weights to domains based on their genome-wide frequencies.
• Constructed a genome-scale human functional network using the developed method.

Main Results:

• The weighted MI method demonstrated superior performance compared to other domain-based network inference techniques.
• The inferred functional networks showed high predictability for biological pathways and phenotypes.
• The genome-scale human functional network identified numerous communities significantly associated with known pathways and diseases.

Conclusions:

• Protein domain co-occurrence analysis using weighted MI is an effective strategy for inferring functional PPIs.
• Domain-based functional networks hold potential for mapping domain-to-pathway and domain-to-phenotype associations.
• This approach can contribute to a deeper understanding of disease mechanisms and the development of targeted therapies.