Thresholds a non sparse adjascency matrix to return the strongest weights only

Source: R/fct_network_inference.R

Without thresholding, we would obtain a fully connected weighted graph from GENIE3, with far too many links to be interpretable. In order build a meaningful network, this weighted adjacency matrix between regulators and targets has to be sparsified, and we have to determine the regulatory weights that we consider significant.

This method is a nice exploratory way to threshold complete networks, but to get more robust and significant results, consider using the DIANE::test_edges() function.

network_thresholding(mat, n_edges)

Arguments

mat

matrix containing the importance values for each target and regulator, as returned by DIANE::network_inference()
n_edges

number of edges top edges to keep in the final network.

Value

igraph object representing the Gene Regulatory Network

Examples

if (FALSE) {
data("abiotic_stresses")
data("regulators_per_organism")
genes <- get_locus(abiotic_stresses$heat_DEGs)
# mat was inferred using the function network_inference
mat <- abiotic_stresses$heat_DEGs_regulatory_links
network <- DIANE::network_thresholding(mat, n_edges = length(genes))
}