AnnotationHub::mapIds() cannot find existing ENSG (GEO supplemental data cross-referenced with ensembl.org)

Question

[mk]

Anyone know why I'm not getting ENSG ids for some of these symbols?

The example below retrieves NA for multiple symbols, including AAED1 whose ENSG is ENSG00000158122.

> library(AnnotationHub)
> library(org.Hs.eg.db)
> library(GEOquery)
> temp <- tempfile()
> download.file(getGEO("GSM4430459")@header$supplementary_file_1,temp)
> genes <- read.csv(temp)$X
> unlink(temp)
> ensids = mapIds(org.Hs.eg.db,
                keys=genes, 
                column="ENSEMBL",
                keytype="SYMBOL",
                multiVals="first")
> ensids["AAED1"] # here is one of the <NA>
AAED1 
NA

Answer 1

[Kevin Blighe]

Hi, a quick check on NCBI Gene reveals that the official symbol for this is PRXL2C, not AAED1. In this way, I would not have expected org.Hs.eg.db (using 'recent' annotation) to have it. However, I can see that EnsDb.Hsapiens.v86 (older version) does [have it]. So, there must have been an annotation change in the recent Ensembl versions. Important to remember that gene annotation is constantly changing.

org.Hs.eg.db

library(org.Hs.eg.db)
select(org.Hs.eg.db,
  keys = 'AAED1',
  column = c('ENSEMBL', 'SYMBOL'),
  keytype = 'SYMBOL')

Error in .testForValidKeys(x, keys, keytype, fks) : 
  None of the keys entered are valid keys for 'SYMBOL'. Please use the keys method to see a listing of valid arguments.

EnsDb.Hsapiens.v86

library(EnsDb.Hsapiens.v86)
select(EnsDb.Hsapiens.v86,
  keys = 'AAED1',
  column = c('GENEID', 'SYMBOL'),
  keytype = 'SYMBOL')

           GENEID SYMBOL
1 ENSG00000158122  AAED1

------------

If we instead check for the official symbol, PRXL2C, in org.Hs.eg.db:

select(org.Hs.eg.db,
  keys = 'PRXL2C',
  column = c('ENSEMBL', 'SYMBOL'),
  keytype = 'SYMBOL')

  SYMBOL         ENSEMBL
1 PRXL2C ENSG00000158122

----------

In situations like this, one can use limma's alias2SymbolTable() to help retrieve all aliases for your genes.

limma::alias2SymbolTable('AAED1', species = 'Hs')
[1] "PRXL2C"

This simple example also highlights why it's better to use Ensembl or Entrez gene IDs for analyses.

Kevin

[mk]

Thanks, Kevin! That worked perfectly. As always, your clear and thoughtful answer is much-appreciated.

Yeah, I definitely think it's easier to work in unique identifiers and usually only convert to symbol for reporting. Then I happen to need this published data and figure I'll just use a GEO supplementary file to "save time"... Ends up taking 10x longer than just re-quantifying their raw SRA data in Salmon...

[mk]

the alias2SymbolTable approach you mentioned would also be useful in the following situation :

• the GEO data set is human transcriptomic sequencing and access to the SRA is controlled
• the authors have posted the quantitated data in their supplement (public access)
• like the above case, the symbols correspond to an unknown annotation version