Get symbols (or any other ids) corresponding to Ensembl gene ids

query_biomart.R

# 1. Install biomart.

source("http://bioconductor.org/biocLite.R")
biocLite("biomaRt")

# 2. Load biomart.

library(biomaRt)

# 3. Get symbols for Ensembl IDs

ensembl_ids = c(
  "ENSG00000243485", "ENSG00000237613", "ENSG00000186092", "ENSG00000238009"
)

mart = useMart(biomart = "ensembl", dataset = "hsapiens_gene_ensembl")

dat = getBM(
  values = ensembl_ids,
  filters = c("ensembl_gene_id"),
  attributes = c("ensembl_gene_id", "external_gene_name", "description"),
  mart = mart
)

write.table(
  dat,
  file = "biomart.tsv",
  quote = FALSE,
  sep = "\t",
  row.names = FALSE
)

# 4. Example: change the row names for a dataframe from ensembl ids to gene names.

# Read the data.
d <- read.delim("counts.tsv.gz", row.names = 1)

# Query biomart.
library(biomaRt)
ensembl_ids <- rownames(d)
mart <- useMart(biomart = "ensembl", dataset = "hsapiens_gene_ensembl")
x <- getBM(
  mart = mart,
  values = ensembl_ids,
  filter = c("ensembl_gene_id"),
  attributes = c("ensembl_gene_id", "external_gene_name")
)

# Create a dictionary from ensembl id to gene name.
ens_to_gene <- as.character(x$external_gene_name)
names(ens_to_gene) <- as.character(x$ensembl_gene_id)

# Some ensembl ids do not map to any gene names.
gene_names <- ens_to_gene[rownames(d)]
gene_names[is.na(gene_names)] <- rownames(d)[is.na(gene_names)]

# Multiple ensembl ids map to a single gene name, so average those ensembl ids.
library(reshape2)
library(data.table)
mean_by <- function(dat, xs) {
  dat <- data.table(dat)
  dat$agg_var <- xs
  dat <- melt(dat, id.vars = "agg_var")
  dat <- dcast.data.table(
    dat, agg_var ~ variable, value.var = "value",
    fun.aggregate = mean, na.rm = TRUE
  )
  rownames(dat) <- dat$agg_var
  dat[ , agg_var := NULL]
  dat
}

# Write the output.
dd <- mean_by(d, gene_names)
write.table(dd, "out.tsv", quote = FALSE, sep = "\t", col.names = NA)