Skip to contents

Identify privileged communities based on a given diffusion vector

community_detection.Rd

Detects privileged communities after a diffusion based on seed nodes.

Usage

community_detection(
  object,
  community_params = params_community_detection(),
  seed = 42L,
  .verbose = FALSE,
  .max_iters = 100L
)

Arguments

object

NetworkDiffusions object. The underlying class NetworkDiffusions().

community_params

List. Parameters for the community detection within the reduced network, see params_community_detection(). A list with the following items:

  • max_nodes - Integer. Number of maximum nodes per community. Larger communities will be recursively subclustered.

  • min_nodes - Integer. Minimum number of nodes per community.

  • min_seed_nodes - Integer. Minimum number of seed genes that have to be found in a given community.

  • initial_res - Float. Initial resolution parameter for the Leiden clustering.

  • threshold_type - String. One of c("prop_based", "pval_based"). You can chose to include a certain proportion of the network (like in the original paper) with the highest diffusion scores, or use p-values based on permutations. Defaults to "prop_based".

  • network_threshold - Float. The proportion of the network to include. Used if threshold_type = "prop_based".

  • pval_threshold - Float. The maximum p-value for nodes to be included. Used if threshold_type = "pval_based".

seed

Random seed.

.verbose

Controls the verbosity of the function.

.max_iters

Controls how many iterations shall be tried for the sub-clustering. To note, in each iteration of the sub-clustering, the resolution parameter is increased by 0.05, to identify more granular communities within the sub communities.

Value

The class with added diffusion community detection results (if any could be identified with the provided parameters).

Examples

# privileged communities in the heated part of the network
set.seed(42)
g <- igraph::sample_pa(15, directed = FALSE)
edges <- data.table::setDT(igraph::as_data_frame(g))[, `:=`(
  from = sprintf("node_%i", from),
  to = sprintf("node_%i", to)
)]
object <- NetworkDiffusions(edges, weighted = FALSE, directed = FALSE)
object <- diffuse_seed_nodes(object, c(node_1 = 1, node_3 = 1), "max")
object <- permute_seed_nodes(object, perm_iters = 100L, .verbose = FALSE)
object <- community_detection(
  object,
  community_params = params_community_detection(
    min_seed_nodes = 0L,
    min_nodes = 2L
  )
)
head(get_results(object))
#>    cluster_id node_id    ks_pval cluster_size seed_nodes_no diffusion_score
#>        <char>  <char>      <num>        <int>         <int>           <num>
#> 1:  cluster_1  node_1 0.00965701            5             1      0.12792857
#> 2:  cluster_1  node_2 0.00965701            5             1      0.24907561
#> 3:  cluster_1  node_5 0.00965701            5             1      0.07379672
#> 4:  cluster_1 node_13 0.00965701            5             1      0.04910154
#> 5:  cluster_1 node_10 0.00965701            5             1      0.05292857
#> 6:  cluster_2  node_3 0.16923077            3             1      0.15054927
#>    seed_node
#>       <lgcl>
#> 1:      TRUE
#> 2:     FALSE
#> 3:     FALSE
#> 4:     FALSE
#> 5:     FALSE
#> 6:      TRUE