Identifying network communities using higher-order structures

Traditional network community detection methods focus on identifying groups of nodes that contain more edges within the group than expected. However, real-world networks often exhibit rich topological structure beyond pairwise relationships, which is better characterized by motifs or graphlets. Thus, it is important to understand communities in terms of higher-order connectivity patterns. To this end, we introduce a graphlet-based community detection method that considers partitioning networks according to their high-order connectivity. Our approach provides a systematic way to obtain higher-order communities and offers a more descriptive view of network organization. When applied to a number of biological networks, we find that it detects functionally relevant groups that are not found by edge-based community detection.