October Recap: Network Analysis

IGRAPH object in R

Generally, graphs, (a.k.a networks) are made up of vertices and edges, which represent the whole inter-relationship of nodes. The R Package igraph uses a particular class called “IGRAPH” to represent it.

IGRAPH object can be created by various ways, including generating from tranditional R object data.frame and adjacant matrix.

Below is the one example to create a IGRAPH object representing a relationship between five actors.

library(igraph)
library(tidyverse)

# vertices -- actors
actors <- data.frame(
  name=c("Alice", "Bob", "Cecil", "David","Esmeralda"),
  age=c(48,33,45,34,21),
  gender=c("F","M","F","M","F"))

# edges -- relations
relations <- data.frame(
  from=c("Bob", "Cecil", "Cecil", "David","David", "Esmeralda"),
  to=c("Alice", "Bob", "Alice", "Alice", "Bob", "Alice"),
  same.dept=c(FALSE,FALSE,TRUE,FALSE,FALSE,TRUE),
  friendship=c(4,5,5,2,1,1), 
  advice=c(4,5,5,4,2,3)
  )

# make igraph from data.frame
g <- graph_from_data_frame(relations, directed=TRUE, vertices=actors)
g

## IGRAPH 834595d DN-- 5 6 -- 
## + attr: name (v/c), age (v/n), gender (v/c), same.dept (e/l),
## | friendship (e/n), advice (e/n)
## + edges from 834595d (vertex names):
## [1] Bob      ->Alice Cecil    ->Bob   Cecil    ->Alice David    ->Alice
## [5] David    ->Bob   Esmeralda->Alice

During the workshop, Chun also introduced several ways to manipulate IGRAPH, including retrieving vertex and edges, adding/deleting vertex and edge attributes, merging multiple graphs and subsetting graphs, etc. The full material refers to the R markdown script at github.

The measurement and clustering of network

After presenting the basic IGRAPH object, Chun advanced the topic a little bit by introducing several useful metrics to measure a graph.

One cool metric particularly useful for social networks is Transitivity, which measures the how well connected the graph is. Among small work networks, a network with high transivity means that it is very likely that the two friends connected to one given person are also friends to each other. To know more, help("transitivity")

Besides network measurement metrics, Chun also covered several methods in graph clustering, like decomposing graph to inter-connected components (decompose.graph), clustering to small cliques (cliques) and grouping graph to communities/modules (eg.cluster_walktrap).

Network visualization

Visualizing network was highlighted in the workshop. It can be done by default plotting plot(g) to visualize a IGRAPH object with multiple attributes (color, shape and linetype, etc) applied to edges and vertices. Alternatively, one may consider plotting a IGRAPH using ggraph which is very similar ggplot2. It makes creating legends much easier!

Below is an example to use ggraph highlighting the actors gender and friendship between them

library(ggraph)
ggraph(g, layout="kk") +
  geom_edge_link(aes(edge_width=friendship, edge_linetype=same.dept), arrow = arrow(angle=5, length = unit(0.3, "inches"))) +
  geom_node_point(aes(color=gender), size=6) +
  geom_node_text(aes(label=name), nudge_y = -0.1, nudge_x = -0.1) +
  scale_edge_width(range = c(1, 2)) +
  theme_void() 

Showcase and group exercise

At end the workshop, Chun presented the application of network analysis in her work. As a bioinformatics scientist, she is particularly interested in the influence of gene regulatory networks on the genetic susceptibility for childhood diseases. She showed us an example of usig a gene-cis-regulatory-element interaction network to predict target genes for Type 1 diabetes-associated variants. To know more, refer to CHOP spatial and functional genomics center website.

As a finaly activity, we started a group exercise exploring real-world flight data. Do you know that, if you choose to take the same airline company flight for the shortest tranfer, there are 15 ways to go from Philly to Beijing China?

Thank you