Significance-based multi-scale method for network community detection and its application in disease-gene prediction

Area of Science:

• Network Science
• Computational Biology

Background:

• Community detection is crucial in complex networks.
• Existing methods like Modularity face resolution limits, hindering multi-scale analysis.

Purpose of the Study:

• To investigate the Significance measure for community detection.
• To develop a multi-resolution Significance method to address limitations in detecting communities within multi-scale networks.

Main Methods:

• Theoretical analysis of Significance's critical behaviors, including critical number of communities and phase diagrams.
• Development of a generalized, multi-resolution version of the Significance measure.
• Experimental validation on various complex networks.

Main Results:

• Significance demonstrates superior resolution compared to Modularity, especially when community densities differ.
• The multi-resolution Significance effectively identifies communities in multi-scale networks.
• The method relaxes first- and second-type resolution limits.

Conclusions:

• The generalized Significance measure is effective for multi-scale community detection.
• This approach has significant potential in computational biology, particularly for disease-gene identification through multi-scale module mining.