Build a graph object from a STRING database export — build_graph_from

Loads a STRING-formatted protein-protein interaction file (the TSV export with columns `node1, node2, ..., combined_score`) or a user-supplied data frame following the same schema, and converts it into a `tbl_graph` consistent with the rest of the `build_graph_from_*()` family. All STRING evidence channels in the input – `homology`, `coexpression`, `experimentally_determined_interaction`, `automated_textmining`, etc. – are preserved as edge attributes so downstream visualizations can colour or filter edges by evidence type.

Usage

build_graph_from_stringdb(
  stringdb,
  node1_col = "node1",
  node2_col = "node2",
  score_col = "combined_score",
  score_threshold = NULL,
  node_annotation = NULL,
  directed = FALSE,
  module.method = c("Fast_greedy", "Walktrap", "Edge_betweenness", "Spinglass"),
  top_modules = 15,
  seed = 1115
)

Arguments

stringdb: Either a file path (`.tsv` / `.tsv.gz`) or a data.frame already loaded in memory. STRING-style headers (a leading `#` on the first column name) are handled automatically.
node1_col, node2_col, score_col: Character. Column names to use as the edge `from`, `to`, and weight columns. Defaults match STRING's TSV layout (`"node1"`, `"node2"`, `"combined_score"`). Override if you want to weight by an individual evidence channel – e.g. `score_col = "experimentally_determined_interaction"`.
score_threshold: Numeric or `NULL` (default). If non-NULL, edges with `score_col < score_threshold` are dropped. STRING confidence cut-offs are typically `0.4` (medium), `0.7` (high), `0.9` (highest). Pass `NULL` to keep every edge in the input.
node_annotation: Optional data frame; first column must match the protein/gene names appearing in `node1_col` / `node2_col`. If provided, these annotations are attached as vertex attributes via [build_graph_from_df()].
directed: Logical (default: `FALSE`). STRING interactions are undirected by convention.
module.method: Character. Network community detection method. Options: `"Fast_greedy"`, `"Walktrap"`, `"Edge_betweenness"`, `"Spinglass"`.
top_modules: Integer. Number of top-ranked modules to retain; smaller modules are collapsed into `"Others"`.
seed: Integer (default: `1115`). Random seed for reproducibility.

Value

A `tbl_graph` object compatible with [ggNetView()]. Edge attributes include `weight` (the chosen score), `correlation` (signed copy of the score, equal to `weight` for STRING since scores are non-negative), and every other column present in the input (e.g. `combined_score`, `coexpression`, `homology`, `experimentally_determined_interaction`).

Details

STRING's TSV exports use a header line that starts with `#node1`. The leading `#` is stripped automatically; all other column names are kept verbatim.

Examples

if (FALSE) { # \dontrun{
# From a STRING TSV file:
obj <- build_graph_from_stringdb(
  stringdb        = "string_interactions.tsv",
  score_threshold = 0.7
)

# From an in-memory data.frame, no filtering:
df <- data.frame(
  node1          = c("AANAT", "AANAT", "ABCA1"),
  node2          = c("CRY1",  "TPH1",  "SIRT1"),
  combined_score = c(0.608,   0.675,   0.520)
)
obj <- build_graph_from_stringdb(stringdb = df)
} # }