Connected Components

Connected components measure whether all nodes in the knowledge graph can reach each other through paths. For drug repurposing, this matters directly: a drug can only be linked to a disease if they exist in the same connected component. The Largest Connected Component (LCC) is the single giant cluster that, in a well-integrated graph, contains the vast majority of nodes.

For drug repurposing to work, both drugs and diseases must be reachable within the graph. These numbers show what fraction of EC-curated core entities reside in the LCC.

The LCC contains two kinds of nodes: the core entities we care about for drug repurposing (EC-curated drugs and diseases), and all the other biological entities that connect them. Understanding what makes up the non-core majority helps assess whether the graph's connective tissue is biologically meaningful.

The remaining 7,292,163 non-core nodes provide the connective structure between drugs and diseases. The breakdown by biolink parent category shows what kinds of biological entities form this bridge.

Connected components follow a power-law distribution: the LCC is enormous, and everything else is tiny. The table below buckets components by size to show this pattern clearly.

Core entities (EC-curated drugs and diseases) that are not in the LCC cannot participate in path-based drug repurposing. Any minor component containing core entities represents a gap in graph connectivity.

Which knowledge sources contribute edges to the disconnected minor components? If fragmentation is driven by a single source, it may indicate poorly-connected data being brought in. A spread across many sources suggests a more systemic integration gap.

The largest disconnected fragments outside the LCC. Click a row to inspect the nodes and edges within that component.