add more than one edge based on edge attributes using igraph

Question

I would like to know wether, using igraph, it is possible to add edges to a graph depending on the values of different edge attributes.

I have a data.frame, which dput is the following:

df <- structure(list(nodeA = c("CFTR", "CFTR", "CFTR", "CFTR", "CFTR", 
"CFTR"), nodeB = c("CYP7A1", "KRT16", "ABCA3", "SLC22A11", 
"PBK", "ACSM1"), score = c(0.239, 0.24, 0.292, 0.269, 
0.233, 0.168), text = c(129L, 0L, 287L, 246L, 
161L, 155L), mining = c(163L, 241L, 413L, 71L, 92L, 56L), 
experiments = c(0L, 0L, 101L, 0L, 75L, 0L), homologs =c(0L, 
0L, 609L, 0L, 0L, 0L)), .Names = c("nodeA", "nodeB", 
"score", "text", "mining","experiments", 
"homologs"), class = "data.frame", row.names = c(NA, 6L))

I would like to add new edges to the graph (g <- graph.data.frame(df, directed=FALSE) if the value of the edge attributes is different from 0, for example for the edge CFTR--CYP7A1, I would like to add a pair of extra edges (one for the text and another for the mining attributes), I am not interested in score (it is the weight of my the graph)

Answer 1

Here are a couple of ways.

First, rearranging your original data seems a bit easier. Put the data in to long format and assign colours based on the column names.

library(reshape2)
# Data in long format 
# Create graph, with edges add when attributes / columns are greater than zero
m <- melt(df, id=1:2)
m <- m[m$value != 0, ] # keep non-zero values
g <- graph.data.frame(m, directed=FALSE)

# Add colours to the edges
cols = c(score="black", text="blue", mining="green", 
                                  experiments="red", homologs="yellow")
plot(g, edge.color=cols[E(g)$variable])

---

If you want to have the original graph and then add coloured edges for each attribute greater than zero, you can loop through the attributes (edge_attr), and add edges (add_edges) when the condition is met.

We can add the additional edges one at a time (shown for the text attribute)

g <- graph.data.frame(df, directed=FALSE)    
names(edge_attr(g)) # attributes

# Which edges should be added conditioned on text attribute being greater than zero
edge_attr(g, "text")
ats <- edge_attr(g, "text") > 0

#Set edges in graph already to black
E(g)$color <- "black"

# Get head and tail of all edges
ed <- get.edgelist(g)

# subset these by the attribute condition
# combine head and tail nodes in correct format for add_edges
# should be c(tail1, head1, tail2, head2, ..., tailn, headn)
ed <- t(ed[ats, 2:1])

# Add the additional edges
g  <- add_edges(g, ed,  color="blue")
plot(g)

Or add the additional edges in one go

g <- graph.data.frame(df, directed=FALSE)    

# Indicator of attribute > 0
ats <- unlist(edge_attr(g)) > 0

# Repeat the head & tail of each edge
# subset so the same length as relevant attributes
ed <- do.call(rbind, replicate(length(edge_attr(g)), get.edgelist(g), simplify=FALSE))
ed <- t(ed[ats, 2:1])
cols <- rep(c("black", "blue", "green", "red", "yellow"), each=length(E(g)))[ats]

g  <- add_edges(g, ed,  color=cols)
plot(g)