The Promise to Partner. Visualization of networks – analyzing and visualizing connections between (planned) NFDI consortia
In this JupyterNotebook we show you how to visualize and analyze a network. We do this using the example of the consortia that are participating in or have applied to the National Research Data Infrastructure Initiative (NFDI).
As a data basis, we take the Letters of Intent of the respective consortia, in which cooperation partners are named. These mentions are the starting point of our network1.
We do the visualization in a JupyterNotebook or R Notebook/JupyterLab2, so no local installation of R is necessary.
JupyterNotebooks are built in such a way that you have different cells in which you write code (in our case R code).
To run the cell with the code, we can click on "Cell" and "Run Cells" in the menu. Or click with the cursor in the cell and then press SHIFT" and "ENTER" at the same time. You will then see the result of the code displayed directly below the cell.
Before we get started, let's clarify a few terms. A network consists of two components:
- Nodes (circle)
- Edges (line)
Nodes (nodes or vertices) are represented as circles and represent consortia. Edges (edges) are represented as more or less curved lines and emanate from the nodes. They indicate a connection between two nodes.
R is built in such a way that different libraries can be loaded for different functions.
For the network analysis we will use the package igraph3.
With library("igraph") we load the package. Have a great introduction into network analysis with R at
https://kateto.net/network-visualization.
With if (!require("igraph")) install.packages("igraph") we install the package in case it is not available on the current system.
if (!require("igraph")) install.packages("igraph")
library("igraph")Loading required package: igraph
Attaching package: ‘igraph’
The following objects are masked from ‘package:stats’:
decompose, spectrum
The following object is masked from ‘package:base’:
union
The data basis is a two-column listing of the consortia.
The first column (from) contains the consortium whose letter of intent is evaluated. The second column (to) contains the consortium which is named as cooperation partner.
This data is read in by means of the function read.table.
There are three parameters:
header=TRUE(there is a header line in the dataset).sep=","(the values are separated by a comma)text=""(the values themselves are between the quotes)
We pass these values to the self-selected variable NFDI_edges , which is done with the arrow symbol pointing to the left.
The data itself comes from the GitHub gist nfdi-collaborations.csv
# https://gist.github.com/LukasCBossert/9bd04115db3aa9ed974fdc69d3ff227c
NFDI_edges <- read.table(header=TRUE,
sep=",",
text="
from,to
DataPLANT,NFDI4BioDiversity
DataPLANT,NFDI4Chem
GHGA,NFDI4Health
KonsortSWD,BERD@NFDI
KonsortSWD,NFDI4BioDiversity
KonsortSWD,NFDI4Earth
KonsortSWD,NFDI4Health
KonsortSWD,Text+
NFDI4BioDiversity,NFDI4Earth
NFDI4BioDiversity,NFDI4Chem
NFDI4BioDiversity,NFDI4Health
NFDI4BioDiversity,KonsortSWD
NFDI4BioDiversity,DataPLANT
NFDI4Cat,FAIRmat
NFDI4Cat,NFDI4Chem
NFDI4Cat,NFDI4Ing
NFDI4Cat,DAPHNE4NFDI
NFDI4Chem,FAIRmat
NFDI4Chem,NFDI4Ing
NFDI4Chem,NFDI4Cat
NFDI4Chem,DAPHNE4NFDI
NFDI4Chem,PUNCH4NFDI
NFDI4Chem,NFDI4Health
NFDI4Chem,NFDI4BioDiversity
NFDI4Culture,Text+
NFDI4Culture,MaRDI
NFDI4Culture,NFDI4Ing
NFDI4Health,GHGA
NFDI4Health,KonsortSWD
NFDI4Health,NFDI4Chem
NFDI4Health,NFDI4Earth
NFDI4Health,NFDI4BioDiversity
NFDI4Ing,NFDI-MatWerk
NFDI4Ing,FAIRmat
NFDI4Ing,NFDI4Chem
NFDI4Ing,NFDI4Earth
NFDI4Ing,MaRDI
NFDI4Ing,Text+
NFDI4Ing,NFDI4Culture
BERD@NFDI,KonsortSWD
BERD@NFDI,MaRDI
BERD@NFDI,Text+
DAPHNE4NFDI,FAIRmat
DAPHNE4NFDI,NFDI-MatWerk
DAPHNE4NFDI,NFDI4Cat
DAPHNE4NFDI,NFDI4Chem
DAPHNE4NFDI,NFDI4Health
DAPHNE4NFDI,NFDI4Ing
DAPHNE4NFDI,PUNCH4NFDI
FAIRmat,DAPHNE4NFDI
FAIRmat,DataPLANT
FAIRmat,MaRDI
FAIRmat,NFDI-MatWerk
FAIRmat,NFDI4Cat
FAIRmat,NFDI4Chem
FAIRmat,NFDI4DataScience
FAIRmat,NFDI4Ing
FAIRmat,PUNCH4NFDI
MaRDI,BERD@NFDI
MaRDI,FAIRmat
MaRDI,NFDI-MatWerk
MaRDI,NFDI4Cat
MaRDI,NFDI4Chem
MaRDI,NFDI4Ing
MaRDI,PUNCH4NFDI
NFDI-MatWerk,DAPHNE4NFDI
NFDI-MatWerk,DataPLANT
NFDI-MatWerk,FAIRmat
NFDI-MatWerk,MaRDI
NFDI-MatWerk,NFDI4Chem
NFDI-MatWerk,NFDI4DataScience
NFDI-MatWerk,NFDI4Ing
NFDI4DataScience,KonsortSWD
NFDI4DataScience,MaRDI
NFDI4DataScience,NFDI-MatWerk
NFDI4DataScience,NFDI4BioDiversity
NFDI4DataScience,NFDI4Cat
NFDI4DataScience,NFDI4Chem
NFDI4DataScience,NFDI4Culture
NFDI4DataScience,NFDI4Health
NFDI4DataScience,NFDI4Ing
NFDI4DataScience,NFDI4Microbiota
NFDI4Earth,DataPLANT
NFDI4Earth,GHGA
NFDI4Earth,KonsortSWD
NFDI4Earth,NFDI4BioDiversity
NFDI4Earth,NFDI4Cat
NFDI4Earth,NFDI4Chem
NFDI4Earth,NFDI4Culture
NFDI4Earth,NFDI4Health
NFDI4Earth,NFDI4Ing
NFDI4Microbiota,DataPLANT
NFDI4Microbiota,GHGA
NFDI4Microbiota,NFDI4BioDiversity
NFDI4Microbiota,NFDI4Chem
NFDI4Microbiota,NFDI4DataScience
NFDI4Microbiota,NFDI4Health
NFDI4Microbiota,NFDI4Ing
PUNCH4NFDI,DAPHNE4NFDI
PUNCH4NFDI,FAIRmat
PUNCH4NFDI,GHGA
PUNCH4NFDI,MaRDI
PUNCH4NFDI,NFDI4Earth
PUNCH4NFDI,NFDI4Ing
Text+,KonsortSWD
Text+,NFDI4BioDiversity
Text+,NFDI4Culture
Text+,NFDI4Earth
Text+,NFDI4Ing
")Now we can check, if the data has been loaded properly into the variable.
head(NFDI_edges)| from | to | |
|---|---|---|
| <fct> | <fct> | |
| 1 | DataPLANT | NFDI4BioDiversity |
| 2 | DataPLANT | NFDI4Chem |
| 3 | GHGA | NFDI4Health |
| 4 | KonsortSWD | BERD@NFDI |
| 5 | KonsortSWD | NFDI4BioDiversity |
| 6 | KonsortSWD | NFDI4Earth |
So that we can create a network from this dataset,
we have to prepare it and create a igraph graph.4
This is done with the function graph_from_data_frame, to which we pass our dataset.
We also specify that our dataset or network is undirected
(directed=FALSE), that means that the direction as specified by from,to in the dataset
does not matter.
All we care about now is that two consortia are linked.
We pass this information to the variable NFDI_network.
NFDI_network <- graph_from_data_frame(NFDI_edges,
directed = FALSE
)Let us check this again and have a look at the data.
head(NFDI_network) [[ suppressing 19 column names ‘DataPLANT’, ‘GHGA’, ‘KonsortSWD’ ... ]]
6 x 19 sparse Matrix of class "dgCMatrix"
DataPLANT . . . 2 . 1 . . . . . 1 . 1 . 1 1 . .
GHGA . . . . . . . 2 . . . . . . . 1 1 1 .
KonsortSWD . . . 2 . . . 2 . 2 . . . . 1 2 . . 2
NFDI4BioDiversity 2 . 2 . . 2 . 2 . . . . . . 1 2 1 . 1
NFDI4Cat . . . . . 2 . . 1 . 2 2 1 . 1 1 . . .
NFDI4Chem 1 . . 2 2 . . 2 2 . 2 2 1 1 1 1 1 1 .
First, we will set a parameter so that our network always looks the same when the data is the same. This parameter is seed.
We choose an arbitrary number, which may be large.
After that we come to the actual plot.
For this we call the function plot and pass it the variable of our network graph NFDI_network.
For a title we can still specify the parameter main and also we can specify if we want to have a frame around the network with frame=TRUE.
set.seed(9876543)
plot(NFDI_network, # loading data frame
main = "NFDI Network", # adding a title
frame = TRUE # making a frame
)
We see the network of NFDI consortia without any other explicit settings.
The next step we want to do is optimize the layout of the network. Instead of retyping the code for the plot, we will select the content of the last cell, copy and paste it into the next cell.
We'll expand the code this way and work on the network step by step.
There are different algorithms for the layout of networks.
Depending on the data set, sometimes one layout, sometimes the other may be more suitable.
With the layout graphopt5 you usually get a good result.
We pass this value layout.graphopt to the parameter layout.
set.seed(9876543)
plot(NFDI_network, # loading data frame
main = "NFDI Network", # adding a title
frame = TRUE, # making a frame
layout = layout.graphopt, #* better layout options
)
We see the network of NFDI consortia without any other explicit settings.
The network is now already better structured and the distances between the nodes are more harmonious.
If you like, you can try out further layout settings:
layout.circle(circle,circular): Deterministic layout that places the vertices on a circlelayout.drl(drl): The Distributed Recursive Layout algorithm for large graphslayout.fruchterman.reingold(fr): Fruchterman-Reingold force-directed algorithmlayout.kamada.kawai(kk): Kamada-Kawai force-directed algorithmlayout.lgl(large, lgl, large_graph): The Large Graph Layout algorithm for large graphslayout.random(random): Places the vertices completely randomlylayout.reingold.tilford(rt, tree): Reingold-Tilford tree layout, useful for (almost) tree-like graphslayout.sphere(sphere,spherical,circular_3d): Deterministic layout that places the vertices evenly on the surface of a sphere
After we have optimized the arrangement of the nodes, let's tackle the representation of the nodes and edges in the next step.
Various parameters can be adjusted according to your own wishes.
First we want to tackle the color of the nodes.
The parameter is vertex.color and we can specify an HTML color value (for example #ffcc66).6
For the border of the nodes we choose the same color code. The parameter is vertex.frame.color.
The labels of the nodes can also be modified.
The change of the font size is done by the parameter vertex.label.cex, to which we pass the value 0.5.
It is important here that the value is not written in quotes.
This is a relative size and we want the labels to be half the size they were in the previous network.
The color of the label can also be changed.
Quite analogously, the parameter is called vertex.label.color, to which we can also pass the color value as a string, such as "black".
A network consists not only of nodes but also of edges connecting two nodes.
For the color change we need the parameter edge.color, to which we pass for example "#808080".
Besides the color we can also specify the degree of "curvature", which is set with edge.curved and the value 0.1. Again, it is important that no quotes are set.
set.seed(9876543)
plot(NFDI_network, # loading data frame
main = "NFDI Network", # adding a title
frame = TRUE, # making a frame
layout = layout.graphopt, # better layout options
vertex.color = "#ffcc66", #* color of nodes
vertex.frame.color = "#ffcc66", #* color of the frame of nodes
vertex.label.cex = 0.5, #* size of the description of the labels
vertex.label.color = "black", #* color of the description
edge.color = "#808080", #* color of edges
edge.curved = 0.1, #* factor of "curvity"
)
In the previous network representations, all nodes are the same size.
Now we want to add another layer of information and output the node size according to the number of its edges.
We can determine the number of edges per node with the function degree7.
If we pass this function the dataset of the network (degree(NFDI_network)),
then we get the number of edges per node.
We take these values as the size specification for the nodes.
We thus extend the previous code by one line.
The node size is hidden behind the parameter vertex.size and as value we pass the function
degree(NFDI_network).
#data.frame(
degree(NFDI_network) #* calculate number of edges
#) - DataPLANT
- 7
- GHGA
- 5
- KonsortSWD
- 11
- NFDI4BioDiversity
- 13
- NFDI4Cat
- 10
- NFDI4Chem
- 19
- NFDI4Culture
- 7
- NFDI4Health
- 13
- NFDI4Ing
- 19
- BERD@NFDI
- 5
- DAPHNE4NFDI
- 12
- FAIRmat
- 16
- MaRDI
- 14
- NFDI-MatWerk
- 12
- NFDI4DataScience
- 13
- NFDI4Earth
- 15
- NFDI4Microbiota
- 8
- PUNCH4NFDI
- 10
- Text+
- 9
set.seed(9876543)
plot(NFDI_network, # loading data frame
main = "NFDI-Netzwerk", # adding a title
frame = TRUE, # making a frame
layout = layout.graphopt, # better layout options
vertex.color = "#ffcc66", # color of nodes
vertex.frame.color = "#ffcc66", # color of the frame of nodes
vertex.label.cex = 0.5, # size of the description of the labels
vertex.label.color = "black", # color of the description
# color: https://www.w3schools.com/colors/colors_picker.asp
edge.color = "#808080", # color of edges
edge.curved = 0.1, # factor of "curvity"
vertex.size = degree(NFDI_network), #* size of nodes depends on amount of edges
)
We have now introduced a second layer of information into our network and can display the node size in relation to the number of edges.
In the next step, we would like to introduce another component. Until now, it was irrelevant whether a consortium was named first or second in the dataset, i.e., it was irrelevant whether it was the active or the passive collaborator.
Now we would like to consider the distinction in the network. To do this, our graph (network) must be "directed"8.
We introduce a new variable (NFDI_network_directed), which contains the dataset as a directed graph,
which we set with directed = TRUE.
NFDI_network_directed <- graph_from_data_frame(NFDI_edges,
directed = TRUE
)Ok, let us have a look at the data.
head(NFDI_network_directed) [[ suppressing 19 column names ‘DataPLANT’, ‘GHGA’, ‘KonsortSWD’ ... ]]
6 x 19 sparse Matrix of class "dgCMatrix"
DataPLANT . . . 1 . 1 . . . . . . . . . . . . .
GHGA . . . . . . . 1 . . . . . . . . . . .
KonsortSWD . . . 1 . . . 1 . 1 . . . . . 1 . . 1
NFDI4BioDiversity 1 . 1 . . 1 . 1 . . . . . . . 1 . . .
NFDI4Cat . . . . . 1 . . 1 . 1 1 . . . . . . .
NFDI4Chem . . . 1 1 . . 1 1 . 1 1 . . . . . 1 .
We transfer the remaining plot data from the previous cell.
It is now crucial that we pass the new variable with the directed graph to the plot function.
In addition, we also pass the new variable to the degree function.
In the directed network, the curvature of the edges makes it difficult to read.
Therefore we choose the value 0 for edge.curved.
Likewise, the arrowheads should become smaller, which is possible with edge.arrow.size and the relative value 0.5.
set.seed(9876543)
plot(NFDI_network_directed, #<<<<<<< loading data frame
main = "NFDI-Netzwerk", # adding a title
frame = TRUE, # making a frame
layout = layout.graphopt, # better layout options
vertex.color = "#ffcc66", # color of nodes
vertex.frame.color = "#ffcc66", # color of the frame of nodes
vertex.label.cex = 0.5, # size of the description of the labels
vertex.label.color = "black", # color of the description
# color: https://www.w3schools.com/colors/colors_picker.asp
edge.color = "#808080", # color of edges
edge.curved = 0, #<<<<<<<<< factor of "curvity"
vertex.size = degree(NFDI_network_directed), #<<<<<< size of nodes depends on amount of edges
edge.arrow.size = .5, #* arrow size, defaults to 1
)
In the next step, we want to scale the node size according to the inbound edges. The more often a consortium is named as a collaborator, the larger its node will be.
We can modify the function degree for this by adding mode = "in"9.
degree(NFDI_network_directed,
mode = "in")
#data.frame(
degree(NFDI_network_directed,
mode = "in")
#)- DataPLANT
- 5
- GHGA
- 4
- KonsortSWD
- 6
- NFDI4BioDiversity
- 8
- NFDI4Cat
- 6
- NFDI4Chem
- 12
- NFDI4Culture
- 4
- NFDI4Health
- 8
- NFDI4Ing
- 12
- BERD@NFDI
- 2
- DAPHNE4NFDI
- 5
- FAIRmat
- 7
- MaRDI
- 7
- NFDI-MatWerk
- 5
- NFDI4DataScience
- 3
- NFDI4Earth
- 6
- NFDI4Microbiota
- 1
- PUNCH4NFDI
- 4
- Text+
- 4
set.seed(9876543)
plot(NFDI_network_directed, # loading data frame
main = "NFDI Network (<in>)", #<<<<<<<< adding a title
frame = TRUE, # making a frame
layout = layout.graphopt, # better layout options
vertex.color = "#ffcc66", # color of nodes
vertex.frame.color = "#ffcc66", # color of the frame of nodes
vertex.label.cex = 0.5, # size of the description of the labels
vertex.label.color = "black", # color of the description
# color: https://www.w3schools.com/colors/colors_picker.asp
edge.color = "#808080", # color of edges
edge.curved = 0, # factor of "curvity"
vertex.size = degree(NFDI_network_directed,
mode = "in"), #<<<<<< size of nodes depends on amount of edges
edge.arrow.size = .5, # arrow size, defaults to 1
)
Likewise, we can now also display the size of the consortia according to their outgoing edges.
We take the complete cell content from before and only change in to out.
#data.frame(
degree(NFDI_network_directed,
mode = "out")
#)- DataPLANT
- 2
- GHGA
- 1
- KonsortSWD
- 5
- NFDI4BioDiversity
- 5
- NFDI4Cat
- 4
- NFDI4Chem
- 7
- NFDI4Culture
- 3
- NFDI4Health
- 5
- NFDI4Ing
- 7
- BERD@NFDI
- 3
- DAPHNE4NFDI
- 7
- FAIRmat
- 9
- MaRDI
- 7
- NFDI-MatWerk
- 7
- NFDI4DataScience
- 10
- NFDI4Earth
- 9
- NFDI4Microbiota
- 7
- PUNCH4NFDI
- 6
- Text+
- 5
set.seed(9876543)
plot(NFDI_network_directed, # loading data frame
main = "NFDI Network (<out>)", #<<<<<<<< adding a title
frame = TRUE, # making a frame
layout = layout.graphopt, # better layout options
vertex.color = "#ffcc66", # color of nodes
vertex.frame.color = "#ffcc66", # color of the frame of nodes
vertex.label.cex = 0.5, # size of the description of the labels
vertex.label.color = "black", # color of the description
# color: https://www.w3schools.com/colors/colors_picker.asp
edge.color = "#808080", # color of edges
edge.curved = 0, # factor of "curvity"
vertex.size = degree(NFDI_network_directed,
mode = "out"), #<<<<<< size of nodes depends on amount of edges
edge.arrow.size = .5, # arrow size, defaults to 1
)
It is noticeable that some nodes are shrinking and in the table you can see that they have the value 0 for outgoing edges. This is because these are the consortia that were already approved in the first funding round and therefore did not submit a new Letter of Intent. After all, our dataset only considers the Letters of Intent from the second funding round.
The consortia of the first round can therefore only be mentioned as "passive" cooperation partners.
After the previous rounds of network visualization, let's go one step further and analyze the network structure.
As a first step, let's color the nodes or consortia in the colors of the NFDI conference systematics.
How does the NFDI conference systematics come about? Five panels have been set up for the presentations:
- Medicine
- Life Sciences
- Humanities
- Engineering Sciences
- Chemistry/Physics
The applicant consortia were divided among these five groups:10
In the following, we abbreviate Group 4 "Computer Science, Mathematics and Engineering" as "Engineering".
It is noticeable that according to the DFG subject classification system, the natural sciences have been divided between the life sciences, engineering sciences and chemistry/physics, as can be seen in the following Sankey (flow chart).
So all consortia have been assigned to one of these five areas and we now want to show this in the network. We load this classification of the consortia on the conference system in the next cell.
This new record is passed to the variable 'NFDI_nodes'; the first column contains the consortium names, the second column the number from the NFDI-conferencesystematics.
The third column contains the round in which the consortium was approved: 1= 2019, 2= 2020.
The data can be read from the public GitHub gist nfdi-consortia.csv.
# https://gist.github.com/LukasCBossert/ce56ebd0059b4879c7d11c1090118c25
NFDI_nodes <- read.table(header=TRUE,
sep=",",
text="
name,group,round
DataPLANT,2,1
GHGA,1,1
KonsortSWD,3,1
NFDI4BioDiversity,2,1
NFDI4Cat,5,1
NFDI4Chem,5,1
NFDI4Culture,3,1
NFDI4Health,1,1
NFDI4Ing,4,1
BERD@NFDI,3,2
DAPHNE4NFDI,5,2
FAIRmat,5,2
MaRDI,4,2
NFDI-MatWerk,4,2
NFDI4DataScience,4,2
NFDI4Earth,2,2
NFDI4Microbiota,2,2
PUNCH4NFDI,5,2
Text+,3,2
")Now we still have to create a graph dataset from the dataset, which is again done with graph_from_data_frame. What is new is that we now differentiate what is our edge data frame and what is the list with the nodes.
NFDI_network_directed <- graph_from_data_frame(d = NFDI_edges, # d = data frame =~ edges
vertices = NFDI_nodes, #nodes
directed = TRUE) #directedWe set up the coloring based on the year of the funding.
The colum round is considered and depending on the number (1 or 2) the color is chosen.
NFDI_color_year <- c("lightgreen", # 2019 (1)
"lightblue" # 2020 (2)
)
NFDI_color_groups <- NFDI_color_year[
as.numeric(as.factor(
V(NFDI_network_directed)$round))] # <<<< based on the roundGreat, so now we can apply the color system on the network.
set.seed(9876543)
plot(NFDI_network_directed, # loading data frame
main = "NFDI-Network (<Funding year>)", #<<<<<<<< adding a title
frame = TRUE, # making a frame
layout = layout.graphopt, # better layout options
vertex.color = NFDI_color_groups, #<<<<<<<<<< color of nodes
vertex.frame.color = NFDI_color_groups, #<<<<<<<<<< color of the frame of nodes
vertex.label.cex = 0.5, # size of the description of the labels
vertex.label.color = "black", # color of the description
# color: https://www.w3schools.com/colors/colors_picker.asp
edge.color = "#808080", # color of edges
edge.curved = 0, # factor of "curvity"
vertex.size = degree(NFDI_network_directed,
mode = "total"), #<<<<<<<<<<< size of nodes depends on amount of edges
edge.arrow.size = .5, # arrow size, defaults to 1
)
In order to better recognize the node classification on the NFDI conference systematics in the network, we choose a color coding according to the DFG subject systematics (slight adjustment if necessary).
The following values apply
| No. | Designation | HTML color code |
|---|---|---|
| (1) | Medicine | #f5ac9f |
| (2) | Life Sciences | #e43516 |
| (3) | Humanities | #f9b900 |
| (4) | Engineering Sciences | #007aaf |
| (5) | Chemistry/Physics | #6ca11d |
We now pass these color values in sequence to the variable 'NFDI_color_code',
thereby the color values are written into a list.
Using the function c the values are written into a vector,11
with which we can continue.
Now we have to establish the link between the color value and the consortia.
For this we introduce the variable NFDI_color_groups:
Each value from NFDI_color_code has a position number (1-5),
we use this by evaluating the value of the second column of the network graph ($group) as a number and thus passing the color value.
Simplified and from the result, the NFDI conference system number gets the color value that is in the corresponding position in the list of the variable NFDI_color_code.
NFDI_color_code <- c("#f5ac9f", # Medicine
"#e43516", # Life Sciences
"#f9b900", # Humanities
"#007aaf", # Engineering Sciences
"#6ca11d" # Chemistry/Physics
)
NFDI_color_groups <- NFDI_color_code[
as.numeric(as.factor(
V(NFDI_network_directed)$group))]We can again take the code from the previous cell and adapt it.
It is crucial that we specify the variable NFDI_color_groups as value for vertex.color and vertex.frame.color.
We also want to consider and display the entire network with all edges (mode = "total").
What is missing now is a legend so that we can also see what is behind the color coding.
set.seed(9876543)
plot(NFDI_network_directed, # loading data frame
main = "NFDI-Network (<NFDI conference systematics>)", #<<<<<<<< adding a title
frame = TRUE, # making a frame
layout = layout.graphopt, # better layout options
vertex.color = NFDI_color_groups, #<<<<<<<<<< color of nodes
vertex.frame.color = NFDI_color_groups, #<<<<<<<<<< color of the frame of nodes
vertex.label.cex = 0.5, # size of the description of the labels
vertex.label.color = "black", # color of the description
# color: https://www.w3schools.com/colors/colors_picker.asp
edge.color = "#808080", # color of edges
edge.curved = 0, # factor of "curvity"
vertex.size = degree(NFDI_network_directed,
mode = "total"), #<<<<<<<<<<< size of nodes depends on amount of edges
edge.arrow.size = .5, # arrow size, defaults to 1
)
Ok, we want to add a legend now and since we want to define it only once we make it as a function, which we now fill with values:
- First the positioning of the legend, which we want to have
bottomright, then the title (title = "NFDI conference systematics"), now comes the content of the legend, which is controlled by thelegendparameter: For this we again build a list (c()), in which we enter the desired values. col: Withcolwe set the color scheme and we can directly refer to the NFDI color list via the variableNFDI_color_code.pch: We must not forget thepchparameter, because it is used to define the symbol in the legend. With the value20we select a filled circle.bty: Withbtyand the valuenfornowe do without a frame around the legend.cex(socharacter expansion) is again a relative value and we can specify the font size; similarly,pt.cexworks for the legend symbols.
nfdi_plot_legend <- function(){
legend("topleft", # x-position
title = "NFDI conference systematics", # title
legend = c(
"(1) Medicine",
"(2) Life Sciences",
"(3) Humanities",
"(4) Engineering Sciences",
"(5) Chemistry/Physics"
), # the text of the legend
col = NFDI_color_code , # colors of lines and points beside the legend text
pch = 20, # the plotting symbols appearing in the legend
bty = "n", # no frame, the type of box to be drawn around the legend (n=no frame)
cex = .75, # character expansion factor relative to current par("cex").
pt.cex = 2 # expansion factor(s) for the points
)
}Now we add the legend to the plot.
set.seed(9876543)
plot(NFDI_network_directed, # loading data frame
main = "NFDI Network (<NFDI conference systematics>)", #<<<<<<<< adding a title
frame = TRUE, # making a frame
layout = layout.graphopt, # better layout options
vertex.color = NFDI_color_groups, # color of nodes
vertex.frame.color = NFDI_color_groups, # color of the frame of nodes
vertex.label.cex = 0.5, # size of the description of the labels
vertex.label.color = "black", # color of the description
# color: https://www.w3schools.com/colors/colors_picker.asp
edge.color = "#808080", # color of edges
edge.curved = 0, # factor of "curvity"
vertex.size = degree(NFDI_network_directed,
mode = "total"), #<<<<<<<<<<< size of nodes depends on amount of edges
edge.arrow.size = .5, # arrow size, defaults to 1
)
nfdi_plot_legend()
Let us concentrate on only one consortium and display the connection from or to this consortium.
nfdi_plot_group <- function(NFDI_name) {
set.seed(9876543)
nfdi_local_network <- function(NFDI_name) {
plot(NFDI_network_directed,
main = "NFDI Network (<NFDI conference systematics>)", # adding a title
sub = NFDI_name,
frame = TRUE, # making a frame
layout = layout.graphopt, # better layout options
vertex.color = NFDI_color_groups, # color of nodes
vertex.frame.color = NFDI_color_groups, # color of the frame of nodes
vertex.label.cex = 0.5, # size of the description of the labels
vertex.label.color = "black", # color of the description
# color: https://www.w3schools.com/colors/colors_picker.asp
edge.curved = 0.2, # factor of "curvity"
vertex.size = degree(NFDI_network_directed,
mode = "total"), #<<<<<<<<<<< size of nodes depends on amount of edges
edge.arrow.size = .5, # arrow size, defaults to 1
edge.color = with(NFDI_edges,
ifelse(from %in% NFDI_name,"#808080", # grey
ifelse(to == NFDI_name,"#000000", # black
NA)))
)
nfdi_plot_legend()
}
####################################################
### ! important!
### If you want to export the plots,
### you need to create a folder called "img" first.
####################################################
# pdf(paste0("img/network_group_",NFDI_name,".pdf")) # save image as PDF
# nfdi_local_network(NFDI_name) # display image for saving
# dev.off() # close image stream
nfdi_local_network(NFDI_name) # display image in JupyterNotebook
}
nfdi_plot_group("NFDI4Ing")
Here is another consortium and its connections.
nfdi_plot_group("NFDI4Microbiota")
I love loops....
for (name in NFDI_nodes$name){
nfdi_plot_group(name)
}
We have done the network visualization and analysis using only the package 'igraph'. Now you have to save the result, e.g. under "File" --> "Save and Checkpoint". You can also download the JupyterNotebook, there are several formats available.
If you have created the network with the RNoteBook, you can call it up again at any time via the URL and you can make further modifications in the network.
There are other exciting occupations with this network. For example, you can also create an interactive network or display the network as a pie chart. Have a look at the overview on https://www.r-graph-gallery.com/network.html.
Footnotes
-
See also the repository of Dorothea Strecker (https://github.com/dorothearrr/NFDI_Netzwerk), where she has already done a similar visualization and analysis. ↩
-
https://mybinder.org/v2/gh/jupyterlab/jupyterlab-demo/master?urlpath=lab/tree/demo cf. https://bookdown.org/yihui/rmarkdown/notebook.html ↩
-
https://www.dfg.de/download/pdf/foerderung/programme/nfdi/nfdi_konferenz_2020/programm_webkonferenz_2020.pdf ↩
-
https://www.rdocumentation.org/packages/base/versions/3.6.2/topics/c ↩