A graph projection is an in-memory representation of selected graph nodes and relationships, prepared for algorithmic analysis rather than general database exploration. Selection can focus analysis on a protein–protein interaction network or a relevant slice of the SNOMED ontology. section 4.2.2 section 7.7.4
A practical projection pipeline is:
- Mark the nodes participating in the analysis, such as proteins connected by INTERACTS_WITH. section 4.2.2
- Project the selected labels and relationships into memory, specifying relationship orientation when required. section 4.2.2
- Run a graph algorithm on the named projection, then write or stream analytical results. section 4.2.2 section 7.7.4
Listing 4.8 Creating the graph projection in memory
call gds.graph.project(
'ppi-graph', #1
'PPIProtein', #2
{
INTERACTS_WITH: { #3
orientation: 'UNDIRECTED'
}
}
)
The projection query names the in-memory graph, selects PPIProtein nodes, and treats INTERACTS_WITH as undirected for downstream GDS algorithms. section 4.2.2
Two analytical projections
A PPI projection supports WCC, which found 27 connected components among 21,559 proteins, and Louvain, which found 48 communities. A SNOMED projection supports degree analysis for identifying highly connected hub nodes and filtering noisy paths between medical entities. section 4.2.2 section 7.7.4
