ggraph can't select columns for geom_node or geom_node_text aes variables

Question

When I try using variables for geom_node_point or geom_node_text I keep getting errors, no matter what I do. Here's what I was trying at first.

library(igraph)
library(ggraph)
library(tidyverse)
graph3 <- graph_from_data_frame(test, directed = F)
ggraph(graph2, layout = "kk") + 
  geom_edge_link(aes(colour = match, alpha = match)) +  geom_node_point(aes(colour = qcluster, alpha =match)) +geom_node_text(repel=TRUE)#  theme(legend.position = "none")

but this gives me the following error

Error in FUN(X[[i]], ...) : object 'qcluster' not found

So then if I try using more precise instructions to show which columns to use, I get other problems

ggraph(graph3, layout = "kk") + 
geom_edge_link(aes(colour = match, alpha = match)) +  geom_node_point(aes(colour = test$qcluster, alpha =match)) +geom_node_text(repel=TRUE)#  theme(legend.position = "none")

returns the error

Don't know how to automatically pick scale for object of type function. Defaulting to continuous. Error: Aesthetics must be valid data columns. Problematic aesthetic(s): alpha = match. Did you mistype the name of a data column or forget to add after_stat()?

So my questions is, how can I get the columns to work as Aes variables, they seem to work just fine in the geom_edge_link, so I'm really confused on what exactly is going wrong. Especially since manually adding aes variables like strings or numbers seems to work just fine. The test dataset, it's a very small subsection of my data so testing goes easier

query	subject	qcluster	scluster	match	bit_sum
16T1_lib200772_4_13	YP_009910784.1	cr165	cr209	False	57.4
16T1_lib200772_4_17	YP_009910789.1	cr177	cr241	False	131.0
16T1_lib200772_4_17	YP_009910790.1	cr177	cr230	False	57.4
16T1_lib200772_4_23	YP_009910794.1	cr7	cr7	True	69.3
16T1_lib200772_4_82	YP_009910759.1	cr1	cr1	True	92.8
16T1_lib200772_4_83	YP_009910760.1	cr6	cr6	True	79.3

score 1 · Answer 1 · answered Sep 27 '21 at 13:38

To build off of Wolfgang's answer, I've found a way to add vertex information without having to resort to manually applying them. graph3 <- graph_from_data_frame(test, directed = F, vertices = unique(FileNames)) where FileNames is a file with metadata on my graph. If you combine this with Wolfgang's answer you'll get the graph you were looking for

score 0 · Accepted Answer · answered Sep 27 '21 at 09:26

I'm also quite new to ggraph - however, what I found out, that when you create a graph with graph_from_data_frame (or as_tbl_graph when using tidygraph) then only one attribute is created for vertices and all other attributes end up as edge attributes which cannot be accessed by geom_node_.... So, my workaround is to add vertex attributes manually. The second error message you get is due to the fact that match() is a function in R. Here is an example that shows how it could work. In this case I was super lazy and added vertex attributes rather arbitrarily - you need to pick the right attribute for your vertices:

test = data.frame(
        query = c("16T1_lib200772_4_13", "16T1_lib200772_4_17",
                "16T1_lib200772_4_17",  "16T1_lib200772_4_23",
                "16T1_lib200772_4_82",  "16T1_lib200772_4_83"),
        subject = c("YP_009910784.1",   "YP_009910789.1",
                "YP_009910790.1",   "YP_009910794.1",
                "YP_009910759.1",   "YP_009910760.1"),
        qcluster = c("cr165",   "cr177",
                "cr177",    "cr7",
                "cr1",  "cr6"),
        scluster = c("cr209",   "cr241",
                "cr230",    "cr7",
                "cr1",  "cr6"),
        my_match = c(FALSE, FALSE,
                FALSE,  TRUE,
                TRUE,   TRUE),
        bitsum = c(57.4,    131.0,
                57.4,   69.3,
                92.8,   79.3))

library(igraph)
library(ggraph)
library(tidyverse)
#library(tidygraph)
graph3 <- graph_from_data_frame(test, directed = F)

#########################################################
#### quick and dirty addition of a vertex attribute: ####
graph3 <- set_vertex_attr(graph3, "qcluster", value = c("cr165",
                "cr177", "cr177", "cr7",
                "cr1",  "cr6", "cr177", "cr177",
                "cr7",  "cr1", "cr6"))
#########################################################

#graph3 <- as_tbl_graph(graph3) 

ggraph(graph3, layout = "kk") + 
  geom_edge_link(aes(colour = my_match)) +
  geom_node_point(aes(colour = qcluster)) +
  geom_node_text(repel=TRUE)#  theme(legend.position = "none")

Thanks this really points me in the right direction, but how am I supposed to pick the vertex attributes? Simpy selecting test$qcluster tells me that the number of items to replace is not a multiple of replacement length — A foot of cinnamon, Sep 27 '21 at 13:03
One option (assuming that test$query and test$subject don't have duplicates and there is exactly one qcluster that belongs either to test$query or test$qcluster is following (replace code between the comments ##### above): `my_vertices <- vertex_attr(graph3, "name") qcluster_vertex_attr <- unname(sapply(my_vertices, function (x) ifelse(x %in% test$query, test$qcluster[test$query == x], test$qcluster[test$subject == x]))) graph3 <- set_vertex_attr(graph3, "qcluster", value = qcluster_vertex_attr)` — Wolfgang Arnold, Sep 28 '21 at 11:22

ggraph can't select columns for geom_node or geom_node_text aes variables

2 Answers2