Inferring gene-phenotype associations via global protein complex network propagation

Area of Science:

• Genomics
• Systems Biology
• Bioinformatics

Background:

• Phenotypically similar diseases often stem from functionally related genes, indicating a modular genetic organization.
• Protein complexes, crucial for biological functions, reflect the modularity of protein-protein interaction networks.
• Understanding these networks aids in elucidating gene-phenotype associations for diseases.

Purpose of the Study:

• To introduce a novel computational technique, Random Walker on Protein Complex Network (RWPCN), for predicting and prioritizing disease-associated genes.
• To leverage the modular organization of protein complex networks for enhanced gene-phenotype association discovery.
• To improve the accuracy of identifying candidate genes for complex human diseases.

Main Methods:

• Constructed a protein complex network using human protein complexes and interaction data.
• Developed RWPCN to compute associations between protein complexes and disease phenotypes.
• Employed leave-one-out cross-validation to assess prediction performance.
• Applied the method to identify novel disease genes for Breast Cancer and Diabetes.

Main Results:

• RWPCN demonstrated superior performance in predicting and prioritizing disease genes compared to existing methods.
• The technique effectively utilized the modularity of protein complex networks for gene prioritization.
• Successfully predicted novel disease gene associations for selected complex diseases.

Conclusions:

• Exploiting modularity in both disease phenomes and protein interactomes enhances guilt-by-association predictions.
• RWPCN offers a robust framework for interrogating the human phenome-interactome network.
• Protein complex networks provide a more effective basis for disease gene detection than simple protein interaction networks.