Skip to contents

Identify the TF to gene regulation

identify_tf_to_genes.Rd

Generates the TF to gene associations from the importance matrix produced by the tree-based regression models. Two filtering strategies are available:

  • Threshold (method = "threshold"): For each gene, computes mean + n_sd * SD of the importance scores across all TFs and retains only pairs exceeding that threshold. This adapts to the per-gene importance distribution and is less sensitive to differences between learners.

  • Top-k (method = "top_k"): Selects the top k_tfs TFs per gene and/or the top k_genes genes per TF. Both margins can be combined (union). At least one of k_tfs or k_genes must be provided.

Both methods accept an optional min_importance floor.

Usage

identify_tf_to_genes(
  x,
  method = c("threshold", "top_k"),
  k_tfs = NULL,
  k_genes = NULL,
  n_sd = 2,
  min_importance = NULL,
  .verbose = TRUE
)

# S3 method for class 'ScenicGrn'
identify_tf_to_genes(
  x,
  method = c("threshold", "top_k"),
  k_tfs = NULL,
  k_genes = NULL,
  n_sd = 2,
  min_importance = NULL,
  .verbose = TRUE
)

Arguments

x

ScenicGrn object.

method

Character. Either "top_k" or "threshold".

k_tfs

Optional integer. Top TFs per gene (only used when method = "top_k").

k_genes

Optional integer. Top genes per TF (only used when method = "top_k").

n_sd

Numeric. Number of standard deviations above the per-gene mean to use as the threshold (only used when method = "threshold"). Default is 2.

min_importance

Optional numeric in [0, 1]. Absolute minimum importance score for inclusion.

.verbose

Boolean. Controls verbosity of the function.

Value

The ScenicGrn object with a tf_to_gene_results data.table added.

Examples

# keep the three strongest TFs per gene
sc <- demo_single_cells()
grn <- scenic_grn_sc(
  sc,
  tf_ids = sprintf("gene_%02d", 1:5),
  scenic_params = params_scenic(
    min_counts = 1L,
    learner_params = list(n_trees = 20L)
  ),
  .verbose = FALSE
)
grn <- identify_tf_to_genes(
  grn,
  method = "top_k",
  k_tfs = 3L,
  .verbose = FALSE
)
head(get_tf_to_gene(grn))
#>         tf    gene importance
#>     <char>  <char>      <num>
#> 1: gene_01 gene_01  0.4643152
#> 2: gene_03 gene_01  0.3568220
#> 3: gene_04 gene_01  0.1553617
#> 4: gene_02 gene_02  0.4926334
#> 5: gene_03 gene_02  0.2651168
#> 6: gene_04 gene_02  0.1317756

unlink(sc@dir_data, recursive = TRUE, force = TRUE)