Obtaining Random Sequences from Given Taxonomic Grouping

[afitz]

Hello,

I apologize if this question is too simple, I am new to bioinformatics and am trying to completely my first independent project. I am trying to retrieve DNA sequences from a set of random organisms within a given taxonomic group. For example, I want to be able to input "Mammalia" and retrieve subsets of say, 5 mammalian genomes. I have been looking into the NCBI resources including the taxdump files, the Taxonomy database, and RefSeq, but am struggling to put these resources together in order to traverse a taxonomy and retrieve random sequences from different taxonomic levels.

Any hints on how/where to begin would be appreciated so much! Thank you!!

[afitz]

"So you want a program that will parse a database of curated reference genome sequences based on user input, then extract a subset of those genomes from a subset of those reference genomes?"

I don't necessarily need to extract a subset of the genomes - I just want a way to obtain a random subset of the genomes contained in a given taxonomic category. For example, I would want to be able to ask for Bacteria and receive several bacterial genomes from a sequence database. Thanks for your question!

[gringer]

I just want a way to obtain a random subset of the genomes contained in a given taxonomic category. For example, I would want to be able to ask for Bacteria and receive several bacterial genomes from a sequence database.

This sounds like too specific a task for pre-existing code, but that doesn't mean someone else hasn't thought similarly in the past and made their own solution. Traversing the NCBI taxonomy is somewhat difficult, but doable. You'd probably be working of the taxonomy data from here:

ftp://ftp.ncbi.nlm.nih.gov/pub/taxonomy/

In particular, nodes.dmp and names.dmp inside taxdump to get/parse the tree, and gi_taxid_nucl.dmp to get genbank accession ID to taxID mappings. I guess the trick will be to filter those accession IDs to only have full chromosomal (or contig) sequences, rather than subsets of sequence.

Once you have NCBI accession numbers, you can retrieve the IDs and sequences using eSearch and eFetch:

http://www.ncbi.nlm.nih.gov/books/NBK25498/#chapter3.ESearch__ESummaryEFetch

You do need to be a bit careful when extracting tons of sequence with eFetch, because it has a maximum limit on the sequences that it will return in one request (something like 10,000).

Another problem for you will be what you mean by "random". The NCBI taxa aren't very well structured, so you will be getting quite a biased sample (i.e. weighted heavily on the more researched organisms) by picking sequences using a uniform distribution.