SBGNview has collected pathway data and gene sets from the following databases: Reactome, PANTHER Pathway, SMPDB, MetaCyc and MetaCrop. These gene sets can be used for pathway enrichment analysis.
In this vignette, we will show you a complete pathway analysis workflow based on GAGE + SBGNview. Similar workflows have been documented in the gage package using GAGE + Pathview.
Please cite the following papers when using the open-source SBGNview package. This will help the project and our team:
Luo W, Brouwer C. Pathview: an R/Biocondutor package for pathway-based data integration and visualization. Bioinformatics, 2013, 29(14):1830-1831, doi: 10.1093/bioinformatics/btt285
Please also cite the GAGE paper when using the gage package:
Luo W, Friedman M, etc. GAGE: generally applicable gene set enrichment for pathway analysis. BMC Bioinformatics, 2009, 10, pp. 161, doi: 10.1186/1471-2105-10-161
Please see the Quick Start tutorial for installation instructions and quick start examples.
In this example, we analyze a RNA-Seq dataset of IFNg KO mice vs wild type mice. It contains normalized RNA-seq gene expression data described in Greer, Renee L., Xiaoxi Dong, et al, 2016.
The RNA abundance data was quantile normalized and log2 transformed, stored in a “SummarizedExperiment” object. SBGNview input user data (gene.data or cpd.data) can be either a numeric matrix or a vector, like those in pathview. In addition, it can be a “SummarizedExperiment” object, which is commonly used in BioConductor packages.
library(SBGNview)
library(SummarizedExperiment)
data("IFNg", "pathways.info")
count.data <- assays(IFNg)$counts
head(count.data)
wt.cols <- which(IFNg$group == "wt")
ko.cols <- which(IFNg$group == "ko")
ensembl.pathway <- sbgn.gsets(id.type = "ENSEMBL",
species = "mmu",
mol.type = "gene",
output.pathway.name = TRUE
)
head(ensembl.pathway[[2]])
if(!requireNamespace("gage", quietly = TRUE)) {
BiocManager::install("gage", update = FALSE)
}
library(gage)
degs <- gage(exprs = count.data,
gsets = ensembl.pathway,
ref = wt.cols,
samp = ko.cols,
compare = "paired" #"as.group"
)
head(degs$greater)[,3:5]
head(degs$less)[,3:5]
down.pathways <- row.names(degs$less)[1:10]
head(down.pathways)
The abundance values were log2 transformed. Here we calculate the fold change of IFNg KO group v.s. WT group.
ensembl.koVsWt <- count.data[,ko.cols]-count.data[,wt.cols]
head(ensembl.koVsWt)
#alternatively, we can also calculate mean fold changes per gene, which corresponds to gage analysis above with compare="as.group"
mean.wt <- apply(count.data[,wt.cols] ,1 ,"mean")
head(mean.wt)
mean.ko <- apply(count.data[,ko.cols],1,"mean")
head(mean.ko)
# The abundance values were on log scale. Hence fold change is their difference.
ensembl.koVsWt.m <- mean.ko - mean.wt
#load the SBGNview pathway collection, which may takes a few seconds.
data(sbgn.xmls)
down.pathways <- sapply(strsplit(down.pathways,"::"), "[", 1)
head(down.pathways)
sbgnview.obj <- SBGNview(
gene.data = ensembl.koVsWt,
gene.id.type = "ENSEMBL",
input.sbgn = down.pathways[1:2],#can be more than 2 pathways
output.file = "ifn.sbgnview.less",
show.pathway.name = TRUE,
max.gene.value = 2,
min.gene.value = -2,
mid.gene.value = 0,
node.sum = "mean",
output.format = c("png"),
font.size = 2.3,
org = "mmu",
text.length.factor.complex = 3,
if.scale.compartment.font.size = TRUE,
node.width.adjust.factor.compartment = 0.04
)
sbgnview.obj