diffuse-methods: Network diffusion

In cbg-ethz/knockdown: Statistical Analysis of High-Throughput Genetic Perturbation Screens

Description

Propagate the estimated gene effects from a previous analysis over a network using network diffusion. First the estimated effects are normalized and mapped to a given genetic network, for instance a PPI or co-expression network. Then the normalized effects are propagated across the edges of the network using a Markov random walk with restarts. By that the initial ranking of genes (as given by their absolute effect sizes) is re-evaluated and the genes are reordered. Thus network diffusion potentially reduced false negative hits.

Usage

diffuse(obj, graph = NULL, r = 0.5, delete.nodes.on.degree = 0,
  do.bootstrap = FALSE, take.largest.component = TRUE,
  correct.for.hubs = TRUE)

## S4 method for signature 'HMAnalysedPerturbationData'
diffuse(obj, graph = NULL,
  r = 0.5, delete.nodes.on.degree = 0, do.bootstrap = FALSE,
  take.largest.component = TRUE, correct.for.hubs = TRUE)

Arguments

obj	HMAnalysedPerturbationData object
graph	a data.frame or tibble with three columns representing a symbolic edge list. The first two columns contain node ids. The third column has to be called weight and is the non-negative weight of the edge between the two nodes.
r	restart probability of the random walk
delete.nodes.on.degree	delete nodes from the graph with a degree of less or equal than delete.nodes.on.degree
do.bootstrap	run a diffusion on every bootstrap sample in case bootstrap samples are available
take.largest.component	if true takes only the largest connected component of the graph and discards all nodes that are not in the largest component. If false takes the compete graph.
correct.for.hubs	if true corrects for the fact that the stationary distribution of the random walk is biased towards hubs.

Value

returns a NetworkAnalysedPerturbationData object

Examples

 data(rnaiscreen)
 hm.fit <- hm(rnaiscreen)

 graph <- readRDS(system.file(
   "extdata", "graph_small.rds", package = "perturbatr"))
 res <- diffuse(hm.fit, graph=graph, r=1)

cbg-ethz/knockdown documentation built on Feb. 11, 2020, 6:25 p.m.