Counting Organisms Belonging To A Taxon

Question

[Igorfobia]

Hi,

I'd like to get the following information about the genomes available at ftp://ftp.ncbi.nih.gov/genomes/Bacteria/

For each taxonomic rank r (species, genus, etc.) and for each taxid t at rank r, I'd like to know how many organisms have t as ancestor. Actually, in the end I need just this aggregated information: for each feasible feasible n, I'd like to obtain the number of NCBI taxid at rank r having n organisms as descendent in the tree.

With the term "organism", I refer to a complete genome contained in the repository above - that is one for each subfolder. In this way, I will also to take into account when there are two or more organisms whose genome is available and are associated to the same species.

I hope my terminology was not too bad...

Thanks in advance!

Answer 1

[Science_Robot]

You can get this information from the NCBI Taxonomy database: ftp://ftp.ncbi.nih.gov/pub/taxonomy/taxdump.tar.gz

The data is in a tree format. The files you will need are names and nodes. Names is a flat text file containing a name and its corresponding taxid. Nodes is a flat text file containing and taxid and its parent. You can load all of this into an associative array:

{ 'id': [ child_1, child_i, child_n ] }

and count the members given a taxid.

Answer 2

[Pablacious]

You could also use EBI's Ontology LookUp Service

http://www.ebi.ac.uk/ontology-lookup/

to access the NEWT Ontology, which is the same NCBI Taxonomy plus a few addendums from UniProt, but "ontologized". Using the ontology-lookup service's web service you can ask for the children or parents of a particular term. These are the signatures of the methods you could use (in Java):

public Map getTermParents(String termId, String ontologyName);
public Map getTermChildren(String termId, String ontologyName, int distance, int[] relationTypes);
public Map getTermRelations(String termId, String ontologyName);
public Map getChildrenFromRoot(String termId, String ontologyName, Vector childrenIds);

The web service is exposed through a WSDL file, so if you don't like Java (I think they provide a built Java client), you can consume it from your favourite language. And you avoid all the parsing :-). I use myself the OLS web service for other tasks, so I would definitely recommend it.