I’ve been using g:Profiler for GO enrichment for a while. It works great. Bundled with gprofiler2 R package, I can do GO enrichment analysis for many species within R.
Today, I wanted to try the Maize-GAMER GO annotation. And this post is to show how you can use user-defined annotation in clusterProfiler.
See code below.
library(tidyverse)
library(clusterProfiler)
# Load GAMER.
zmaGO <- readr::read_tsv("https://github.com/timedreamer/public_dataset/raw/master/maize.B73.AGPv4.aggregate.gaf.gz", skip=1) %>% select(term_accession, db_object_id)
# Load GO_ID to GO_annotation mapping.
GOname <- readr::read_tsv("https://github.com/timedreamer/public_dataset/raw/master/agriGOv2_GOConsortium_term_v201608.txt.gz", col_names = c("GO","type","name","number")) %>% select(GO, name)
# Run GO enrichment.
ego <- enricher(gene = unique(zmaGO$db_object_id[1:100]),
TERM2GENE = zmaGO, TERM2NAME = GOname)
head(ego)
barplot(ego, showCategory=20)
PS:
- When loading GAMER, there is a warning
Warning: 2 parsing failures.. You can ignore it. - I also tried
go.obofrom http://geneontology.org/docs/download-ontology/. More GOs overlap with GAMER in the AgriGO mapping than this one. So I go with AgriGO (see below for the numbers).
# Some quick check. Skip.
## A few GOs in GAMER are not in the annotation mapping.
## How many GOs in GAMER?
length(unique(zmaGO$term_accession)) # 9336
# length(unique(GOname$GO))
## How many GOs overlap between GAMER and annotation mapping?
length(intersect(unique(zmaGO$term_accession), unique(GOname$GO))) # 9304
- To parse the
go.obofile, I used sgrote/OboToTerm. Works very well.