Difference between revisions of "GMODTools"

Jump to: navigation, search
m (Contact : add cat)
Line 176: Line 176:
[[Category:Database Tools]]
[[Category:Database Tools]]
[[Category:GMOD Components]]
[[Category:GMOD Components]]

Revision as of 19:31, 6 June 2008


Bulkfiles is a GMOD Perl package that generates Fasta, GFF, DNA and other bulk genome annotation files from Chado databases. It works with several FlyBase chado releases, with SGDLite, and has been tested with other chado databases. Once tuned to your project's needs with its organism and site configurations, it can generate public data releases on a regular basis. It produces all the contents needed for a GMOD Standard URL genome data download folder.


These are current primary outputs, which are configurable to suit project needs.

  • DNA files (full chromosomes) in raw and fasta formats
  • GFF (v3) feature files
  • Fasta sequence for each selected feature set, with headers from feature files
  • BLAST Index files (NCBI)
  • Gene Ontology association file
  • GenBank genome submission files

Additional genome outputs can be added with Perl adaptor packages from a Bulkfiles base adaptor object.

Why Use Bulkfiles?

Why use this package rather than using other middleware layers to chado db - chadoxml, chadodbi, bioperl, ? The general logic is

1. dump all chado db features using simple (and quick) sql, to common intermediate table files, and chromosome dna to raw files. The feature info is simple: type, location, name/id, and a few attributes (db_xrefs,..)

2. postprocess these table files to create the various public use formats (the time-consuming and configurable part), organized into per-chromosome files.

Here are some reasons we take this approach:

a. using simple sql to dump all db features to intermediate table allows easy checks that all features get to bulk files

b. simple sql dump is fast, and reliable in getting all mapped features by keeping logic simple

c. process table output in stages - better debugging of steps in process, and can split processing among computers These stages are loosely coupled - one can go back, tweek configurations and get a new output w/o redoing the complete extraction process.

d. convert one common feature table + dna to several output formats in one step, or repeatedly as needed.

e. combine features from several chado dbs, and add other sources that may not be in your chado database.

f. model organism projects need fairly complex and data specific configurations - moving that to config files keeps code reusable.

g. each genome chado database has different policy and choices with respect to feature, vocabulary and other data. A highly configurable tool, with data extraction and correction methods that are separate and tunable is needed to adapt to such variation in genome databases.


Here is a candidate release package for GMODTools:

 curl -O http://eugenes.org/gmod/GMODTools/GMODTools-1.2.zip
 unzip GMODTools*.zip

If you want to try out GMODTools from CVS check out using these commands:

 cvs -d:pserver:anonymous@gmod.cvs.sourceforge.net:/cvsroot/gmod login
 cvs -d:pserver:anonymous@gmod.cvs.sourceforge.net:/cvsroot/gmod co schema/GMODTools

These commands will create a directory named schema, with a directory named GMODTools in it. Note that you don't need to supply a password, and it may be very slow - the SourceForge anonymous CVS server is notoriously overworked.


Bulkfiles has extensive configurations, in a simple XML format. This is both a strength and weakness. The strength is that most aspects for genome data publication, such as feature names, types, aspects of the output format, are under your project's control. The weakness is these need detailed documentation to make it easier to tune your site's configuration.

All of the organism and project-specific logic is in these configuration files, including output documentation, feature controls and naming, file choices.

Once configured for your organism(s) and project database, it is quick to generate new bulk data release files, and link into a collection of public releases. A new release needs only a minor configuration update (release number and date), and can be generated automatically if desired.


These are the main contents of the configuration files

  • sgdbulk.xml : main release configuration for tested sample
This main release xml controls what other configurations are used.
A new release configuration with date, release name should be added as needed.
  • bulkfiles_template.xml : documented template for creating your project/organism configuration
  • site_defaults.xml  : common database and output settings for your site.
  • chadofeatsql.xml  : primary Chado SQL used to extract data from database.
If you have a complex chado database, you may well want to add to or update this.
  • chadofeatconv.xml : logic to convert chado view to public view of features.
What features are to be published, the structure of features, and much of the messy genome detail are included here. It is complex but that is part of the territory with genome databases.
  • per-output configurations like fastawriter.xml, blastfiles.xml, genbanksubmit.xml have some site configurations such as path to NCBI tools that will need attention for proepr use.

Sample Use

Load a genome chado db to Postgres database:

 curl -O http://sgdlite.princeton.edu/download/sgdlite/sgdlite.sql.gz
 createdb sgdlite
 (gunzip -c sgdlite.sql.gz | psql -d sgdlite -f - ) >& log.load

Extract bulk files from database:

 cd GMODTools
 perl -Ilib bin/bulkfiles.pl -conf sgdbulk -make

If your test fails, pleas re-run with the -debug option and send the result log file to the developer contact below.

 perl -Ilib bin/bulkfiles.pl -conf sgdbulk -make -debug >& gmodtools-debug.log

Sample Output Data Folder

Output data folder looks like this. It is suited to linking to a public web server for data downloading, e.g. the GMOD Standard_URL.

 Example.txt  README.html  blast/       fasta/       gff/         tmp/
 HEADER.html  Release.txt  dna/         fff/         tables/

See the tables/ folder for summary tables. The contents including web HEADER, README and others are configured with the conf/bulkfiles/ configurations.

Sample Run Log

microbe%  perl -Ilib bin/bulkfiles.pl -conf sgdbulk -make
- Setting GMOD_ROOT=/bio/argos/gmod/GMODTools
Config: title = SGD Lite; date = 20051129;  from conf/bulkfiles/sgdbulk.xml
Config: title = Site Default settings for GMODTools;  from conf/bulkfiles/site_defaults.xml
Config: title = Species abbreviations; date = 20051129;  from conf/bulkfiles/organisms.xml
Config: title = Bulkfiles fileset definitions; date = 20040821;  from conf/bulkfiles/filesets.xml
Config: title = Chado Feature mapping info; date = 20040821;  from conf/bulkfiles/featuresets.xml
 missing data dir data/genomes/Saccharomyces_cerevisiae
Config: title = Chado DB SQL; date = 20051129;  from conf/bulkfiles/chadofeatsql.xml
Automaking feature_table files
Config: title = Blast index writer; date = 20040821;  from conf/bulkfiles/blastfiles.xml
Config: title = Summary Tables; date = 20051217;  from conf/bulkfiles/tablewriter.xml
Config: title = Gbrowse conf generator; date = 20040826;  from conf/bulkfiles/gbrowseconf.xml
Config: title = Genome Web docs; date = 20051225;  from conf/bulkfiles/genomeweb.xml
Changed 'current' release symlink to data/genomes/Saccharomyces_cerevisiae/sgdlite_2005_08_23; ok=1
Bulkfiles done. result=overviews:, fff+gff=146661, fasta=19849, blast=14, tables=ok,
Bulkfiles are located at data/genomes/Saccharomyces_cerevisiae/sgdlite_2005_08_23

GenBank genome round-trip test

See also this brief GMODTools_TestCase that describes how to load a GenBank genome to Chado then regurgitate it via Bulkfiles as a GenBank submission file set.


Bulkfiles is mostly a self-contained Perl package. It uses a few BioPerl parts plus XML::Simple for configuration files.

  • Postgres and GMOD Chado database.
  • Basic Perl tool set used for other GMOD packages.


Version 1.2 (2008 May) IN PROGRESS

  • adding GenbankSubmitWriter, that creates a genome submission (new or update) for NCBI.

There now is enough of a GMODTools framework for dumping Chado genome databases to Genbank Submit format that it likely will save effort for those who need to do this job. This is open source and collaborators are welcome to add code here

 esp. lib/Bio/GMOD/Bulkfiles/GenbankSubmitWriter.pm and conf/bulkfiles/genbanksubmit.xml

THe above code is packaged at http://eugenes.org/gmod/GMODTools/ as GMODTools-1.2.zip

Here is sample Bulkfiles outputs from DrosMel CHR_4 and AnoGam CHR_X http://insects.eugenes.org/genome/Drosophila_melanogaster/dromel_20080512/ http://insects.eugenes.org/genome/Anopheles_gambiae_str._PEST/anogam_20080511/

Version 1.1 (2007 October) adds these features and corrections:

  • No chromosome/scaffold/golden_path files. This change is needed to handle partially assembled genomes with many (1000s to 100,000s) of scaffolds. Flag no_csomesplit=1 to use this (should become default).
  • Gene Ontology association file, see go_association tags in configurations
  • Validate main variables in chado database: ${golden_path}, ${seq_ontology}. This step, on now by default, checks that database contains configured values. If failed, db is inspected for real values.
  • Miscellany bugs cured and configuration updates, e.g., tables/overview now again active.

See Also

  • XORT offers an alternative approach to bulk uploads and downloads from a Chado database.
  • Sample_Chado_SQL gives an overview of the feature and other Chado table SQL commands used for Bulkfiles output. Examples here can help you learn the contents of a database created by someone else, and are useful for GMODTools configuration.