Assessment of network module identification across complex diseases.

Area of Science:

• Bioinformatics
• Systems Biology
• Genomics

Background:

• Numerous bioinformatics methods exist for simplifying complex biological networks into subnetworks or modules.
• The comparative performance of these methods in identifying disease-relevant modules across different network types is not well understood.

Purpose of the Study:

• To comprehensively assess and compare the performance of various module identification methods.
• To evaluate methods across diverse biological networks, including protein-protein interaction, signaling, gene co-expression, homology, and cancer-gene networks.
• To identify top-performing algorithms for detecting disease-associated modules.

Main Methods:

• The 'Disease Module Identification DREAM Challenge' was established as an open competition.
• Seventy-five module identification methods were assessed.
• Predicted network modules were tested for association with complex traits and diseases using 180 genome-wide association studies.

Main Results:

• A robust assessment of 75 module identification methods was conducted.
• Top-performing algorithms were identified, demonstrating the ability to recover complementary trait-associated modules.
• Most identified modules correspond to core disease-relevant pathways, many of which include therapeutic targets.

Conclusions:

• The challenge provides biologically interpretable benchmarks, tools, and guidelines for molecular network analysis.
• This work advances the study of human disease biology through improved network analysis techniques.
• The findings highlight the importance of network module identification in discovering disease mechanisms and potential therapeutic targets.