DiffNet is an R package that is aimed at identifying differences in network topology. The algorithm takes as input omics data corresponding to two different conditions, infers a network of interactions between all variables in the data set, and then computes the difference in (inferred) network topology between the two conditions.
The easiest way to install the DiffNet package is using the devtools package:
library(devtools)
devtools::install_github("diogocamacho/diffnet")
Using the example data set provided with the library, we will show how to run the algorithm. First, we'll start by loading the library:
library(diffnet)
Next, let's select a random set of samples to classify as Group 1 and a second set as Group 2. We have 120 samples in the data set, so each group will have exactly 60 samples:
set.seed(1234)
g1 <- sample(x = seq_along(1:120), size = 60, replace = FALSE)
g2 <- setdiff(seq_along(1:120), g1)
With the groups defined, we can now run the algorithm. We will run it using the default parameter for correlation type (Spearman) and correlation threshold (0.75):
D <- diff_top(data = dataset$expression_data, group1 = g1, group2 = g2)
Running this will generate a tibble with 27,409 rows, where each row corresponds to an edge in the network and the corresponding difference score. Using the gene annotations provided with the example data:
D %>%
dplyr::mutate(x = dataset$gene_annotations$symbol[x]) %>%
dplyr::mutate(y = dataset$gene_annotations$symbol[y]) %>%
dplyr::select(., x, y, change_type, total_score)