Why this Network plot doesn't keep the nodes from same group together in R?

Question

I have dataframe nodes with the group information like below:

dput(nodes)

structure(list(id = c("AC006329.1", "AC027796.4", "AC111170.2", 
"AC111170.3", "AC138207.4", "AP000695.2", "CDK15", "COL14A1", 
"COL15A1", "DDX11-AS1", "FOXP4-AS1", "IFNG-AS1", "ITGB2-AS1", 
"LINC00944", "LINC01213", "LINC01395", "MIR155HG", "MSTRG.108144", 
"MSTRG.110466", "MSTRG.11483", "MSTRG.130624", "MSTRG.134576", 
"MSTRG.147129", "MSTRG.180740", "MSTRG.180762", "MSTRG.24184", 
"MSTRG.9061", "SERHL"), label = c("AC006329.1", "AC027796.4", 
"AC111170.2", "AC111170.3", "AC138207.4", "AP000695.2", "CDK15", 
"COL14A1", "COL15A1", "DDX11-AS1", "FOXP4-AS1", "IFNG-AS1", "ITGB2-AS1", 
"LINC00944", "LINC01213", "LINC01395", "MIR155HG", "MSTRG.108144", 
"MSTRG.110466", "MSTRG.11483", "MSTRG.130624", "MSTRG.134576", 
"MSTRG.147129", "MSTRG.180740", "MSTRG.180762", "MSTRG.24184", 
"MSTRG.9061", "SERHL"), group = structure(c(1L, 6L, 6L, 6L, 4L, 
3L, 4L, 4L, 4L, 1L, 1L, 2L, 2L, 2L, 5L, 5L, 2L, 4L, 2L, 1L, 6L, 
5L, 5L, 3L, 3L, 3L, 1L, 5L), levels = c("blue", "brown", "cyan", 
"green", "purple", "red"), class = "factor")), row.names = c(NA, 
-28L), class = "data.frame")

And the edges look like below:

here is the edges.csv file

Using nodes and edges I used the below code to create the network plot:

When I used like below, the network plot was constantly moving and doesn't stop at all.

visNetwork(nodes, edges)

I also tried adding visPhysics(stabilization = FALSE) to above line, but didn't work.

So, I tried like below with igraph layout:

visNetwork(nodes, edges) %>%
  visIgraphLayout()

This doesn't keep the nodes from same group together. Can anyone please tell me how to keep the nodes from same group / color together?

It should look like below:

that is an example plot, I found online...here it is `https://www.r-bloggers.com/2019/06/interactive-network-visualization-with-r/` I showed this here just to say that I want my plot should look like this. — beginner, Jul 24 '23 at 09:07
I think what you're looking for is a *force directed layout* (instead of the regular layout). The documentation of the package you're using probably has a section on graph layouts in which you'll find the desired function under keywords such as *force directed* or popular algorithms like "Fruchtermann-Reingold" or "Kamada-Kawai". See e.g. https://igraph.org/r/doc/layout_with_fr.html for igraph — I_O, Jul 24 '23 at 09:38
Okay, but nothing changed with `layout_with_fr`. I used this way `visNetwork(data1, data5) %>% visIgraphLayout(layout = "layout_with_fr")` — beginner, Jul 24 '23 at 11:36
You have a complete graph. Every vertex is connected to every other. Visualizing it this way is quite pointless ... All the information here is in the edge weights, so you might want to look for a way to indicate these. First, you need a graph layout which takes weights into account and makes high-weight edges short. `layout_with_fr` is like this. I don't know `visIgraphLayout` and what it needs to consider weights. — Szabolcs, Jul 24 '23 at 20:45
Second, it makes sense to show the weights somehow. You can remove weak connections. Or show the weight through the edge thicknesses. — Szabolcs, Jul 24 '23 at 20:46
If you could show me that with an example, it would be great. Thank you. — beginner, Jul 25 '23 at 09:46

Allan Cameron · Accepted Answer · 2023-07-24T15:26:55.420

I usually find it easier to customize a network graph using the tidygraph and ggraph ecosystem. Your graph has a lot more edges than the given example, so it might take some tweaks to get the edges looking as you want them.

library(tidygraph)
library(ggraph)

# Match colours in given example:
colormap <- c(red = '#fa3233', blue = '#3d85e9',
  green = '#4cb30f', purple = '#7b29ef',
  cyan = '#e02aef', brown = '#ffa500')

fillmap <- c(red = '#fb7e81', blue = '#97c2fc',
             green = '#7be142', purple = '#ad85e4',
             cyan = '#ea7cf3', brown = '#ffff00')

tbl_graph(nodes, 
          cbind(edges, colour = nodes$group[match(edges$from, nodes$id)])) %>%
  ggraph(layout = 'igraph', algorithm = 'fr') +
  geom_edge_diagonal(aes(color = colour, width = weight)) +
  geom_node_circle(aes(r = 1, color = group, fill = group)) +
  coord_equal() +
  scale_edge_color_manual(values = colormap) +
  scale_color_manual(values = colormap) +
  scale_fill_manual(values = fillmap) +
  scale_edge_width_continuous(range = c(0.1, 5)) +
  theme_void() +
  theme(legend.position = 'none')

Edit

To adjust the layout so that the groups are not too close together, and to label the nodes in a legible way, you could do:

tbl_graph(nodes, 
          cbind(edges, colour = nodes$group[match(edges$from, nodes$id)])) %>%
  ggraph(layout = 'igraph', algorithm = 'fr', weights = (edges$weight)^0.2) +
  geom_edge_diagonal(aes(color = colour, width = weight)) +
  geom_node_label(aes(label = label, fill = group, color = group), size = 3,
                  label.r = unit(0.5, "lines")) +
  geom_node_text(aes(label = label), size = 3) +
  coord_equal(clip = 'off') +
  scale_edge_color_manual(values = colormap) +
  scale_color_manual(values = colormap) +
  scale_fill_manual(values = fillmap) +
  scale_edge_width_continuous(range = c(0.1, 5)) +
  theme_void() +
  theme(legend.position = 'none')

Awesome. But how can I name the nodes? I want to label the nodes and Is there a way to make the groups bit far..not too close — beginner, Jul 24 '23 at 14:29

ThomasIsCoding · Answer 2 · 2023-07-24T15:23:26.777

3

This is what I have tried with igraph

g <- edges %>%
    mutate(
        color1 = with(nodes, as.character(group)[match(from, id)]),
        color2 = with(nodes, as.character(group)[match(to, id)])
    ) %>%
    mutate(color = ifelse(color1 == color2, color1, "gray")) %>%
    mutate(width = ifelse(color == "gray", weight, weight * 100)) %>%
    graph_from_data_frame(directed = FALSE) %>%
    set_vertex_attr(name = "color", value = with(nodes, as.character(group)[match(names(V(.)), id)])) %>%
    set_vertex_attr(name = "label", value = with(nodes, label[match(names(V(.)), label)])) %>%
    set_vertex_attr(name = "size", value = 10) %>%
    set_edge_attr(name = "curved", value = TRUE)

and then plot it

plot(g, layout = layout_with_fr)

edited Jul 24 '23 at 15:23

answered Jul 24 '23 at 14:07

ThomasIsCoding

96,636
9
24
81

cool. but the names are `group` names, instead I need labels for the nodes. And how can I stretch the plot? – beginner Jul 24 '23 at 14:27
@beginner see my update – ThomasIsCoding Jul 24 '23 at 14:46

Why this Network plot doesn't keep the nodes from same group together in R?

2 Answers2