Co-occurence networks using presence/absence data R

Question

I am trying to make a co-occurence network graph for my presence/absence data of bacteria species but am unsure how to go about with it. I'm hoping to end up with something like this enter image description herewhere each species is linked to another species if they are both present in the same patient, with a larger circle for higher frequency species. I originally tried using widyr and tidygraph packages but I'm not sure if my data set is compatible with them enter image description here, as it has the patients as columns and the individual species as rows. Preferably I would like to know what packages/code I could use that would work with my data set, or how I could change my data set to work with these packages.

Answer 1

You can use a matrix cross product to get a co-occurrence matrix. Then it is simple to convert the adjacency matrix into a graph with igraph package. Try this:

library(igraph)

# Create fake data set
# rows = patients
# cols = species
set.seed(2222)
df <- matrix(sample(c(TRUE, FALSE), 50, replace = TRUE), 5)
colnames(df) <- letters[1:10]

# Generate co-occurrence matrix with crossproduct
co_mat <- t(df) %*% df

# Set diagonal values to 0
diag(co_mat) <- 0

# Assign dim names
dimnames(co_mat) <- list(colnames(df), colnames(df))

# Create graph from adjacency matrix
# ! edge weights are equal to frequency of co-occurrence
g <- graph_from_adjacency_matrix(co_mat, mode = "upper", weighted = TRUE)

# Assign nodes weight equal to species frequency
g <- set.vertex.attribute(g, "v_weight", value = colSums(df))

plot(g, vertex.size = V(g)$v_weight * 5 + 5, edge.width = E(g)$weight * 5)

Here is our fake data

         a     b     c     d     e     f     g     h     i     j
[1,]  TRUE  TRUE  TRUE  TRUE FALSE FALSE FALSE FALSE  TRUE FALSE
[2,]  TRUE FALSE FALSE FALSE  TRUE  TRUE  TRUE FALSE  TRUE FALSE
[3,] FALSE  TRUE FALSE FALSE  TRUE FALSE  TRUE FALSE FALSE FALSE
[4,] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE  TRUE  TRUE
[5,] FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE  TRUE  TRUE FALSE

And here is a result:

Answer 2

Like Istrel, I would also recommend igraph. May a second solution with ggplot..

library(ggnetwork)
library(ggplot2)
library(igraph)

#sample data:
set.seed(1)
mat <- matrix(rbinom(50 * 5, 1, 0.1), ncol = 15, nrow = 100)

# This is not necessary for the example data. But in your case, if you want  species as nodes you have to do a transpose: 
#mat <- t(mat)

#### Optional! But usually there are often "empty cases" which you might want to remove: 
# remove 0-sum-columns
mat <- mat[,apply(mat, 2, function(x) !all(x==0))] 
# remove 0-sum-rows
mat <- mat[apply(mat, 1, function(x) !all(x==0)),]

# transform in term-term adjacency matrix
mat.t <- mat  %*% t(mat)

##### calculate graph 
g <- igraph::graph.adjacency(mat.t,mode="undirected",weighted=T,diag=FALSE)

# calculate coordinates (see https://igraph.org/r/doc/layout_.html for different layouts)
layout <- as.matrix(layout_with_lgl(g))

p<-ggplot(g, layout = layout, aes(x = x, y = y, xend = xend, yend = yend)) +
  geom_edges( color = "grey20", alpha = 0.2, size = 2) + # add e.g. curvature =  0.15 for curved edges
  geom_nodes(size =  (centralization.degree(g)$res +3) , color="darkolivegreen4", alpha = 1) +
  geom_nodes(size =  centralization.degree(g)$res , color="darkolivegreen2", alpha = 1) +
  geom_nodetext(aes(label = vertex.names), size= 5) +
  theme_blank()
p

enter image description here

Use ggplot aesthetics:

# calculate degree:
V(g)$Degree <- centralization.degree(g)$res

p<-ggplot(g, layout = layout, aes(x = x, y = y, xend = xend, yend = yend)) +
  geom_edges( color = "grey20", alpha = 0.2, size = 2) + # add e.g. curvature = 0.15 for curved edges
  geom_nodes(aes(size =  Degree) , color="darkolivegreen2", alpha = 1) +
  scale_size_continuous(range = c(5, 16)) +
  geom_nodetext(aes(label = vertex.names), size= 5) +
  theme_blank()
p