Constructs a network of complexes where edges represent similarity.
Usage
buildComplexNetwork(
complexes,
enrichments,
mode = "combined",
similarityMethod = "jaccard",
alpha = 0.75,
nCores = NULL,
chunkSize = 1000,
verbose = TRUE
)Arguments
- complexes
A named list of protein complexes.
- enrichments
A named list of enrichment results (from `runComplexEnrichment`).
- mode
Edge weight mode: "compositional", "functional", or "combined". Defaults to "combined".
- similarityMethod
The metric for similarity. Defaults to **"jaccard"**. **Warning:** Using "overlap" is discouraged as it tends to collapse diverse hierarchies into single blobs.
- alpha
Numeric (0-1). Weight given to Compositional Similarity in "combined" mode. Defaults to **0.75** (favoring physical structure).
- nCores
Number of cores for parallel processing.
- chunkSize
Size of processing chunks.
- verbose
Logical.
Details
**Systems Biology Rationale:** To generate a diverse and physically meaningful landscape, this function defaults to prioritizing **Compositional Similarity** (shared proteins) over functional similarity.
- **Composition (Protein Identity)** is treated as the "Physical Truth". - **Function (Enrichment)** is treated as an attribute.
The `alpha` parameter controls this balance. The new default (0.75) gives 75 (like "Cell Cycle") from artificially pulling distinct physical complexes into a single cluster, while still allowing functionally related complexes to drift closer than unrelated ones.