In montilab/hypeR-next: Network Enrichment

knitr::opts_chunk$set(message=FALSE, warning=FALSE, comment="")
devtools::load_all(".")
library(igraph)
library(hypeR)
library(dplyr)
library(visNetwork)

data(ig)
set.seed(1)
layout <- igraph::layout_with_graphopt(ig, 
                                       start=NULL, 
                                       niter=1000,
                                       charge=0.005,
                                       mass=30, 
                                       spring.length=0,
                                       spring.constant=1, 
                                       max.sa.movement=5)

Enrichment of a single community

Enrichment is performed independently for each community. Rather than performing enrichment just on the community members, we want to account for neighboring members outside but within close proximity to the community. Therefore we perform network propagation seeded with the community neighbors and performing enrichment on highly traversed portions of the graph.

library(hypeR)
genesets <- hypeR::msigdb_gsets("Homo sapiens", "H", clean=TRUE)

data <- enrich_community(ig, community="16", genesets, restart=0.5, plotting=TRUE)

par(mar=c(0,0,0,0))
ig.c <- data$ig
plot(ig.c,
     vertex.size=4,
     vertex.color=V(ig.c)$pcolor,
     vertex.label=NA,
     edge.width=1,
     layout=layout)

After traversing the network, we can define a ranked signature for this community - sort of like an extended signature - based on the most traversed nodes in the network.

head(igraph::as_data_frame(ig.c, what="vertices"))

signature.ranked <- V(ig.c)$p
names(signature.ranked) <- V(ig.c)$symbol
head(sort(signature.ranked, decreasing=TRUE))

This happens implicitly within this function.

hyp.c <- data$hyp

hyp.c$plots$`Dna Repair`

hypeR::hyp_dots(hyp.c)

Restart probability

The most important parameter will be the restart probability of the random walks. This tells the algorithm how long the walks should be and is the probability of the random walk restarting from the community. In otherwords, this will define how fuzzy your extended signature for a given community will be.

par(mar=c(0,0,0,0), mfrow=c(2,3))
for (restart in c(0.75, 0.5, 0.25, 0.1, 0.05, 0)) {
    data <- enrich_community(ig, community="16", genesets, restart=restart)
    ig.c <- data$ig
    plot(ig.c,
         vertex.size=4,
         vertex.color=V(ig.c)$pcolor,
         vertex.label=NA,
         edge.width=1,
         layout=layout)
    title(paste0("Restart = ", restart), cex.main=2, font.main=2, line=-3)
}

Enrichment of a all communities

One way to visualize enrichment of the network would be to repeat the above process for each community

communities <- names(table(V(ig)$community))
lhyp <- mapply(function(community) {
    data <- enrich_community(ig, community=community, genesets, restart=0.5)
    return(data$hyp)
}, communities, SIMPLIFY=FALSE, USE.NAMES=FALSE)
names(lhyp) <- paste0("C", communities)

mhyp <- hypeR::multihyp$new(data=lhyp)
hypeR::hyp_dots(mhyp, merge=TRUE, top=50, fdr=0.01)

Although we can also visualize the enrichment on the graph itself, which will give some perspective on why some communities close to each other might be enriched for similar pathways.

ig.e <- enrich_communities_static(ig, genesets, restart=0.5, fdr=0.01, top=1)

par(mar=c(0,0,0,0))
set.seed(1234)
plot(ig.e,
     vertex.size=5,
     vertex.label=V(ig.e)$enrichment,
     vertex.color=adjustcolor(V(ig.e)$color, alpha.f=0.6),
     vertex.frame.color=adjustcolor("#000000", alpha.f=0),
     vertex.label.family="Helvetica",
     vertex.label.color="black",
     vertex.label.font=2,
     vertex.label.cex=0.8,
     vertex.label.dist=runif(vcount(ig.e), -0.25, 0.25),
     edge.color=adjustcolor("#000000", alpha.f=0.2),
     layout=layout)

Interactive community enrichment

We have interactive versions of the same functionality that allows one to compactly fit more data into the network.

vn.e <- enrich_communities_interactive(ig, genesets, restart=0.5, fdr=0.25, top=3, val="fdr")

Here you will get the original network but hovering over the nodes reveals community enrichment information.

visNetwork::visIgraphLayout(vn.e, layoutMatrix=layout)

or

visNetwork::visIgraphLayout(vn.e, layout="layout_nicely", randomSeed=1)

Meta interactive community enrichment

You can optionally create a low-dimensional meta-network of the communities with enrichment information where the edges between communities are weighted by the number of shared edges. We also set a minimum shared edge threshold to highlight communities with the largest overlap.

vn.e <- enrich_communities_interactive_meta(ig, genesets, restart=0.5, fdr=0.05, top=3, mse=3, val="fdr")

visNetwork::visIgraphLayout(vn.e, layout="layout_nicely", randomSeed=1)

montilab/hypeR-next documentation built on March 11, 2023, 11:50 a.m.