Random Network Analysis

The script enables the creation of a random network, which will be uploaded on Cytoscape for visualization and manipulation. Moreover, the script will analyse the basic features of the network such as the mean degree distribution of the nodes, and it will produce an HTML report file with statistics data and plots for further exploration. The file README contains a detailed description of the code, while the file fastscript.R provides the code as a single command to launch it directly on R. The file additional_features offers more options for the creation and exploration of the network.

Stage 0: R Packages

The packages required for this script can be summarized in three categories:

1. Packages needed for the creation of the network, its analysis and its transfer to Cytoscape
2. Packages needed for the plotting of the data extracted from the network
3. Packages needed for the creation of the HTML report

Below you can find the necessary packages. Please ensure that they are all installed before launching the script. In case you miss some, you can find everything you need within the installation script.

# Packages installation
install.packages("igraph")
install.packages("ggplot2")
install.packages("htmltools")
install.packages("htmlwidgets")
if (!require("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

BiocManager::install("RCy3")

# Packages needed for the script
library(igraph)
library(ggplot2)
library(htmltools)
library(RCy3)
library(htmlwidgets)

Stage 1: Network Creation

Now that all the packages are installed, we can proceed to the creation of the network. The package Igraph can provide different types of random networks. For this script, we will use the erdos.renyi.game. In case at the beginning you prefer to have at your disposal a network that remains constant, you can provide a seed for reproducibility.

# Set a seed for reproducibility. You can choose a number of your preference
set.seed(106)

# Produce a random Network and visualize it
graph <- erdos.renyi.game(200, p = 0.05, directed = FALSE)
plot(graph, main = "Network")

With this script, the network will be immediately available and displayed. As you can see, most of the nodes are overlapping and, as such, it is difficult to visually interpret the network. The package Igraph provides a number of layouts that can help to better analyse the network that we are working with. For better visualization of the nodes the layout style graphopt is recommended.

# Modify the layout of the Network for better visualization, and create a PNG file
layout <- layout.graphopt(graph)
plot(
    graph,
    layout = layout,
    main = "Network_graphopt_style",
    vertex.label.cex = 0.8,
    vertex.size = 8,
)
#Create a PNG file of the stylized plot
png("Network_Graphopt_Style.png", width = 1000, height = 1000)
plot(
    graph,
    layout = layout,
    main = "Network_graphopt_style",
    vertex.label.cex = 0.8,
    vertex.size = 8,
)
dev.off()
# Assign HTML coordinates to the stylized Network for the HTML report
network_graphopt_style_dependency <- htmltools::htmlDependency(
    "network_graphopt_style", "1.0.0", src = "network_graphopt_style.png", script = FALSE,
    stylesheet = FALSE
)
network_graphopt_style_html <- sprintf('<div><img src="%s" alt="network graphopt style"></div>', network_graphopt_style_dependency$src)

The script will modify the layout of the network for a better visualization, and it will save the image produced as a PNG file in the work directory. The parameters for which we provide numbers are used for the modification of the size of the nodes and the etichettes. These can be modified for any need. All we need for Stage 1 is to upload the network on Cytoscape. This is easily done with this script. WARNING: make sure that Cytoscape is open before launching this script or it will not work. In case Cytoscape is not installed, here you can find the link for the download.

#Dowload link for Cytoscape
https://cytoscape.org/download.html

# WARNING: Cytoscape must be open before launching the script
# Connect to Cytoscape and upload the Network
cytoscapePing()
createNetworkFromIgraph(graph, title = "Network", collection = "Maastricht_Assignment")

Stage 2: Analysis of the Network

This script will directly extract the information for a standard network analysis directly from the list that composes the graph. The information is not useful as it is, but it will be used for the creation of the HTML report file.

# Extract the information needed for the analysis of the network
degree_info <- degree(graph) # Number of edges of each node
closeness_info <- closeness(graph) # Closeness to the center of each node
betweenness_info <- betweenness(graph) # Crucial level of connectivity between nodes 
clustering_info <- transitivity(graph) # Tendencies of each node to form a group

Stage 3: Report

With the information obtained from the last script, we can finally create a report file for our network. Before the compilation of the report, we need to create plots useful for the visualization of the data. Below you can find the script for each of them.

# Create the plot of the Degree level
data_deg <- data.frame(Node = 1:length(degree_info), Degree = degree_info)
degree_graph <- ggplot(data_deg, aes(x = Degree)) +
    geom_histogram(binwidth = 1, fill = "blue", color = "black", alpha = 0.7) +
    labs(x = "Degree", y = "Frequency") +
    theme(axis.title=element_text(size=9)) +
    ggtitle("Degree Distribution") +
    theme(plot.title = element_text(hjust = 0.5))
degree_graph  
ggsave("degree_graph.png", degree_graph, device = "png", width = 5, height = 3)
# Create a plot of the Betweeness level
data_bet <- data.frame(Node = 1:length(betweenness_info), Betweenness = betweenness_info)
mean_betweenness <- mean(betweenness_info)
betw_graph <- ggplot(data_bet, aes(x = Node, y = Betweenness)) +
    geom_bar(stat = "identity", fill = "blue", alpha = 0.7) +
    geom_hline(yintercept = mean_betweenness, linetype = "dashed", color = "red", size = 1) +
    labs(x = "Node", y = "Betweenness") +
    theme(axis.title=element_text(size=8)) +
    ggtitle("Betweenness Distribution with Mean") +
    theme_minimal() +
    theme(plot.title = element_text(hjust = 0.5))
betw_graph
ggsave("betw_graph.png", betw_graph, device = "png", width = 5, height = 3)

# Assign HTML coordinates to the plots for the HTML report
degree_graph_dependency <- htmltools::htmlDependency(
    "degree_graph", "1.0.0", src = "degree_graph.png", script = FALSE,
    stylesheet = FALSE
)
betw_graph_dependency <- htmltools::htmlDependency(
    "betw_graph", "1.0.0", src = "betw_graph.png", script = FALSE,
    stylesheet = FALSE
)
degree_graph_html <- sprintf('<div><img src="%s" alt="Degree Graph"></div>', degree_graph_dependency$src)
betw_graph_html <- sprintf('<div><img src="%s" alt="Betw Graph"></div>', betw_graph_dependency$src)

The script will not only create PNG files of the plots but it will also make an HTML address for each of them to add them to the report file. With all the information and the plots ready, all we need is the final report. This is the script needed to obtain it. The style of the report can be modified by changing parameters between the 'style' set. The report contains also a short explanation of the data obtained.

# Produce the HTML report file
report <- paste0(
    "<style>",
    "  body { font-family: 'Arial', sans-serif; font-size: 28px; color: #333; }",
    "  h1, h2 { color: #008080; }",
    "  p { line-height: 1.6; }",
    "  .highlight { background-color: #FFFF00; }",
    "  .center { text-align: center; }",
    "</style>",
    "<h1>Network Analysis Report</h1>",
    network_graphopt_style_html,
    "<h2>Basic Information</h2>",
    "<p>Number of Nodes: ", vcount(graph), " (the number of nodes in the network.)</p>",
    "<p>Number of Edges: ", ecount(graph), " (the number of edges connecting the nodes in the network.)</p>",
    "<h2>Degree Distribution</h2>",
    "<p>Mean Degree: ", mean(degree_info), " (average of the degrees of the nodes in the graph. High values = more connected network.)</p>",
    degree_graph_html,
    "<p>Standard Deviation of Degree: ",  sd(degree_info), " (standard deviation of node degrees. High values = heterogeneity in the degree distribution.)</p>",
    "<h2>Closeness Centrality</h2>",
    "<p>Mean Closeness Centrality: ", mean(closeness_info), " (average of closeness centrality values across all nodes. High values = nodes in the network are closer to each other in terms of the shortest path length.)</p>",
    "<h2>Betweenness Centrality</h2>",
    "<p>Mean Betweenness Centrality: ", mean(betweenness_info), " (average betweenness centrality across all nodes in the graph. High values = nodes in the network have a more crucial role in connecting different parts.)</p>",
    betw_graph_html,
    "<h2>Clustering Coefficient</h2>",
    "<p>Global Clustering Coefficient: ", clustering_info, " (degree to which nodes in a graph tend to cluster together. High values = higher tendency for nodes in the network to form clusters or groups)</p>"
)

writeLines(report, "network_analysis_report.html")

With this script you will obtain a report file filled with the information needed, as well as plots and network images. To open the HTML file, you can use your standard web browser, and an internet connection is not needed. However, if you prefer to stay in R, this final script will show the result of the analysis.

# Produce a report file for visualize the result in R
report_data <- c(
    "Network Analysis Report",
    "Number of Nodes:",vcount(graph),"(the number of nodes in the network.)",
    "Number of Edges:",ecount(graph),"(the number of edges connecting the nodes in the network.)",
    "Mean Degree:",mean(degree_info),"(average of the degrees of the nodes in the graph. High values = more connected network)",
    "Standard Deviation of Degree:",sd(degree_info),"(standard deviation of node degrees. High values = heterogeneity in the degree distribution.)",
    "Mean Closeness Centrality:",mean(closeness_info),"(average of closeness centrality values across all nodes. High values = nodes in the network are closer to each other in terms of the shortest path length.)",
    "Mean Betweenness Centrality:",mean(betweenness_info),"(average betweenness centrality across all nodes in the graph. High values = nodes in the network have a more crucial role in connecting different parts.)",
    "Global Clustering Coefficient:",clustering_info,"(degree to which nodes in a graph tend to cluster together. High values = higher tendency for nodes in the network to form clusters or groups)"
)
cat(report_data, sep = "\n")