reduce_solution_network: reduce_solution_network

In saezlab/COSMOS: COSMOS (Causal Oriented Search of Multi-Omic Space)

reduce_solution_network

Description

Reduces a solution network based on a decoupling analysis of upstream and downstream gene expression, by filtering out edges that do not meet a consistency threshold, and limiting the network to a certain number of steps from upstream input nodes.

Usage

reduce_solution_network(
  decoupleRnival_res,
  meta_network,
  cutoff,
  upstream_input,
  RNA_input = NULL,
  n_steps = 10
)

Arguments

decoupleRnival_res	A data frame resulting from the decoupleRnival function.
meta_network	A network data frame containing signed directed prior knowledge of molecular interactions.
cutoff	The consistency threshold for filtering edges from the solution network.
upstream_input	A named vector with up_stream nodes and their corresponding activity.
RNA_input	A named vector containing differential gene expression data.
n_steps	The maximum number of steps from upstream input nodes to include in the solution network.

Value

A list containing the solution network (SIF) and an attribute table (ATT) with gene expression data.

Examples

# Example input data
upstream_input <- c("A" = 1, "B" = -1, "C" = 0.5)
downstream_input <- c("D" = 2, "E" = -1.5)
meta_network <- data.frame(
  source = c("A", "A", "B", "C", "C", "D", "E"),
  target = c("B", "D", "D", "E", "D", "B", "A"),
  interaction = c(-1, 1, -1, 1, -1, -1, 1)
)
RNA_input <- c("A" = 1, "B" = -1, "C" = 5, "D" = 0.7, "E" = -0.3)

# Run the decoupleRnival function to get the upstream influence scores
upstream_scores <- decoupleRnival(upstream_input, downstream_input, meta_network, n_layers = 2, n_perm = 100)

# Reduce the solution network based on the upstream influence scores
reduced_network <- reduce_solution_network(upstream_scores, meta_network, 0.4, upstream_input, RNA_input, 3)

# View the resulting solution network and attribute table
print(reduced_network$SIF)
print(reduced_network$ATT)

saezlab/COSMOS documentation built on Sept. 17, 2023, 1:22 p.m.