Issues In Mapping Uniprot Ids To Entrez Ids

Question

[mtyler.jason]

Hi all,

I am currently trying to use the uniprot.org site's protein id to entrez id mapping service. I have around 2714 human proteins. However, when I map them to entrez ids, it gives me matches for only 790 proteins which is much lesser than I had anticipated. Are there better services to match uniprot ids to entrez ids?

Answer 1

[Hamish]

By "uniprot ids" do you mean UniProtKB entry names (e.g. BRCA1_HUMAN) or accessions (e.g. P38398)?

Since UniProtKB entry names are subject to change, you may need to map any entry names you are using to accessions before attempting to perform further mappings.

Also do you mean Entrez Gene identifiers or NCBI GI numbers?

If the former, then:

• See the "UniProt, GeneID, RefSeq mapping: how does it work?" FAQ entry for details of the limitations of the Entrez Gene (GeneID) mappings.
• Try using the NCBI E-utilities or the Entrez web interface to query Entrez Gene with your UniProtKB accessions to get the associated Entrez Gene entries which cross-reference those UniProtKB entries.

If you want GI numbers then you can do much the same thing, but instead of querying Entrez Gene you need to query the 'Protein' database (which includes all the sequence data in UniProtKB), and will return the NCBI version of the UniProtKB entry, including the GI number assigned to the current protein sequence.

You could also try mapping your UniProtKB accessions (or entry names) through UniParc to RefSeq. Since this mapping is based on sequence identity, instead of the UniProtKB cross-reference annotations (which do not exist for some organisms), it will give you RefSeq entries which have the same sequence. You can then look-up these RefSeq identifiers in Entrez Gene to get the best coverage possible for a set of sequence entries. It may be possible to go further by including older sequence versions in the mapping, however these are typically updated to correct sequencing errors, and so should be treated with caution.

Answer 2

[miquelduranfrigola]

It's quite surprising that you get such a small coverage. You are using SwissProt Uniprot ACs, right?

Anyway, if you are a R user, biomaRt Bioconductor's package is very useful.