Getting genes from pathway databases

Gene Ontology

The following SQL query gets all gene names annotated to Gene Ontology terms in Humans.

SELECT          DISTINCT T.acc, T.name, GPR.symbol, DBX.xref_dbname, DBX.xref_key
FROM            term T
INNER JOIN      graph_path GP
    ON          term1_id = T.id
INNER JOIN      association A
    ON          A.term_id = GP.term2_id
    AND         A.is_not = 0
INNER JOIN      gene_product GPR
    ON          GPR.id = A.gene_product_id
INNER JOIN      species S
    ON          S.id = GPR.species_id
    AND         S.genus = 'Homo'
    AND         S.species = 'sapiens'
INNER JOIN      dbxref DBX
    ON          DBX.id = GPR.dbxref_id
;

For example, you can run this query in GOOSE or in locally downloaded copy of the Gene Ontology database.

Note: Many more example queries can be found here.

KEGG

The Kyoto Encyclopedia of Genes and Genomes has an FTP site.  It seems to consist of text files and images in custom formats – I found it pretty hard to understand.  However, hidden in there are files containing pathway ids and associated gene ids — for example, here’s is the one for humans.  (I learned this from the GenomeNet support team, who also said

“Column 2 is the list of KEGG gene IDs and Column 3 is that of representative gene names.  KEGG GENES is a collection of gene catalogs generated from publicly available resources, mostly NCBI RefSeq.  Thus, gene names are got from NCBI RefSeq.”

And much indebted I am to them, too.

Reactome

The easiest way to get this information form Reactome seems to be the Reactome Biomart.  (Personally, I’d much rather run a database query, because I’m of that sort of persuasion, but this does not seem possible in a simple way with the Reactome database.  My guess – and it is only a guess – is that Reactome’s object-relational mapping is designed to be manipulated using an object-oriented code framework, and that transitive membership is not computable using the database directly.)