MCODE

MCODE stands for Molecular Complex Detection, a clustering algorithm used to identify densely connected regions within large biological networks, especially protein–protein interaction (PPI) networks. It was introduced in 2003 by researchers Bader and Hogue as a Cytoscape plug-in to help infer putative molecular complexes from high-throughput network data. The method has since become a widely cited tool in network biology for exploring functional modules in biological interaction networks.

The algorithm operates by assigning a weight to each node that reflects local connectivity density within its neighborhood. Nodes with high weights serve as seeds for cluster formation. Clusters are grown by expanding from these seeds to neighboring nodes that meet predefined density criteria, continuing until the cluster can no longer be extended. After initial clustering, optional post-processing steps may be applied to refine results. These refinements typically include a haircut, which removes loosely connected peripheral nodes, and fluff, which can add neighboring nodes that are well connected to many cluster members.

Typical parameters used in MCODE include a degree cutoff, which filters out low-degree nodes during weight calculation, and a node density cutoff, which influences cluster growth. The output consists of sets of clusters that are interpreted as candidate protein complexes or functional modules within the network. Users commonly analyze these clusters for functional enrichment to gain biological insights.

MCODE remains a foundational tool in network analysis, frequently implemented as part of Cytoscape workflows. Its simplicity and focus on dense subgraphs make it a baseline method for modular structure discovery in PPI networks, though results can be sensitive to network quality and parameter choices.