MOBAS: identification of disease-associated protein subnetworks using modularity-based scoring.

Area of Science:

• Computational Biology and Bioinformatics
• Genomics and Systems Biology
• Network Medicine

Background:

• Network-based analyses, particularly protein-protein interaction (PPI) subnetworks, are crucial for interpreting genome-wide association studies (GWAS) in a functional context.
• Existing methods for scoring disease-associated subnetworks often integrate disease association and network connectivity but may yield non-significant or parameter-dependent results.
• Current scoring schemes can produce arbitrarily large subnetworks or require manual tuning, limiting their statistical robustness and biological interpretability.

Purpose of the Study:

• To develop a novel, parameter-free scoring scheme for identifying disease-associated functional modules from GWAS data.
• To integrate the statistical significance of both network connectivity and disease association into a unified scoring function.
• To assess the performance of the proposed scheme compared to existing methods using GWAS datasets for type II diabetes and psoriasis.

Main Methods:

• Proposed a parameter-free scoring scheme that evaluates the disease association of interactions between pairs of gene products.
• Incorporated statistical significance of both network connectivity and disease association into the scoring function.
• Applied the scoring scheme to GWAS datasets for type II diabetes (T2D) and psoriasis (PS) and compared results with commonly used methods.

Main Results:

• Subnetworks identified by commonly used methods may not pass rigorous statistical significance tests after multiple hypothesis testing correction.
• The proposed scoring scheme successfully identifies highly significant subnetworks enriched with biologically relevant proteins.
• The method demonstrates reproducibility across different cohorts and robustly recovers relevant subnetworks even with reduced sampling rates.

Conclusions:

• The novel parameter-free scoring scheme offers a statistically robust and biologically meaningful approach to functional interpretation of GWAS data.
• This method overcomes limitations of existing scoring schemes, providing more reliable identification of disease-associated functional modules.
• The findings highlight the importance of statistically rigorous network analysis for uncovering complex disease mechanisms and potential therapeutic targets.