Robin Browaeys 2019-01-17
To determine signaling paths between a ligand and target of interest, we look at which transcription factors are best regulating the target genes and are most closely downstream of the ligand (based on the weights of the edges in the integrated ligand-signaling and gene regulatory networks). Then, the shortest paths between these transcription factors and the ligand of interests are determined and genes forming part in this path are considered as important signaling mediators. Finally, we look in our collected data source networks for all interactions between the ligand, signaling mediators, transcription factors and target genes. This allows to both prioritize signaling mediators and check which of all collected data sources support the ligand-target predictions of interest.
For this analysis, you need to define:
- one or more ligands of interest
- one or more target genes of interest
In this vignette, we will demonstrate how to infer signaling paths between a CAF-ligand (CAF = cancer-associated fibroblast) of interest and some of its top-predicted p-EMT target genes. The output of this analysis can be easily imported into Cytoscape for exploration of the networks.
First, we will load the necessary packages and networks to infer signaling paths between ligand and target genes of interest.
library(nichenetr)
library(tidyverse)
weighted_networks <- readRDS(url("https://zenodo.org/record/7074291/files/weighted_networks_nsga2r_final.rds"))
ligand_tf_matrix <- readRDS(url("https://zenodo.org/record/7074291/files/ligand_tf_matrix_nsga2r_final.rds"))
lr_network <- readRDS(url("https://zenodo.org/record/7074291/files/lr_network_human_21122021.rds"))
sig_network <- readRDS(url("https://zenodo.org/record/7074291/files/signaling_network_human_21122021.rds"))
gr_network <- readRDS(url("https://zenodo.org/record/7074291/files/gr_network_human_21122021.rds"))As an example, we will infer signaling paths between the CAF-ligand
TGFB2 and its top-predicted p-EMT target genes SERPINE1 and COL1A1. For
better visualization of edge weights, we will also normalize edge
weights to make them comparable between signaling and gene regulatory
interactions (minmax_scaling = TRUE).
ligands_oi <- "TGFB2" # this can be a list of multiple ligands if required
targets_oi <- c("SERPINE1","COL1A1")
active_signaling_network <- get_ligand_signaling_path(ligands_all = ligands_oi,
targets_all = targets_oi,
weighted_networks = weighted_networks,
ligand_tf_matrix = ligand_tf_matrix,
top_n_regulators = 4,
minmax_scaling = TRUE)
graph_min_max <- diagrammer_format_signaling_graph(signaling_graph_list = active_signaling_network,
ligands_all = ligands_oi, targets_all = targets_oi,
sig_color = "indianred", gr_color = "steelblue")
# To render the graph in RStudio Viewer, uncomment following line of code
# DiagrammeR::render_graph(graph_min_max, layout = "tree")
# To export/draw the svg, you need to install DiagrammeRsvg
graph_svg <- DiagrammeRsvg::export_svg(DiagrammeR::render_graph(graph_min_max, layout = "tree", output = "graph"))
cowplot::ggdraw() + cowplot::draw_image(charToRaw(graph_svg))
We will now look which of the collected data sources support the interactions in this network.
data_source_network <- infer_supporting_datasources(signaling_graph_list = active_signaling_network,
lr_network = lr_network, sig_network = sig_network, gr_network = gr_network)
head(data_source_network)
## # A tibble: 6 × 5
## from to source database layer
## <chr> <chr> <chr> <chr> <chr>
## 1 NFKB1 COL1A1 regnetwork_source regnetwork regulatory
## 2 NFKB1 COL1A1 trrust trrust regulatory
## 3 NFKB1 COL1A1 omnipath_ABC omnipath regulatory
## 4 NFKB1 SERPINE1 trrust trrust regulatory
## 5 NFKB1 SERPINE1 omnipath_ABC omnipath regulatory
## 6 SMAD3 COL1A1 regnetwork_source regnetwork regulatoryFor information of all mentioned data sources in the source column (link to the website of the database, etc), see Data source information
Export the following to e.g. Cytoscape for exploration of the networks
output_path <- ""
write_output <- FALSE # change to TRUE for writing output
# weighted networks ('import network' in Cytoscape)
if(write_output){
bind_rows(active_signaling_network$sig %>% mutate(layer = "signaling"),
active_signaling_network$gr %>% mutate(layer = "regulatory")) %>%
write_tsv(paste0(output_path,"weighted_signaling_network.txt"))
}
# networks with information of supporting data sources ('import network' in Cytoscape)
if(write_output){
data_source_network %>% write_tsv(paste0(output_path,"data_source_network.txt"))
}
# Node annotation table ('import table' in Cytoscape)
specific_annotation_tbl <- bind_rows(
tibble(gene = ligands_oi, annotation = "ligand"),
tibble(gene = targets_oi, annotation = "target"),
tibble(gene = c(data_source_network$from, data_source_network$to) %>% unique() %>% setdiff(c(targets_oi,ligands_oi)) %>% intersect(lr_network$to %>% unique()), annotation = "receptor"),
tibble(gene = c(data_source_network$from, data_source_network$to) %>% unique() %>% setdiff(c(targets_oi,ligands_oi)) %>% intersect(gr_network$from %>% unique()) %>% setdiff(c(data_source_network$from, data_source_network$to) %>% unique() %>% intersect(lr_network$to %>% unique())),annotation = "transcriptional regulator")
)
non_specific_annotation_tbl <- tibble(gene = c(data_source_network$from, data_source_network$to) %>% unique() %>% setdiff(specific_annotation_tbl$gene), annotation = "signaling mediator")
if(write_output){
bind_rows(specific_annotation_tbl, non_specific_annotation_tbl) %>%
write_tsv(paste0(output_path,"annotation_table.txt"))
}