iRefIndex MITAB2.6 Parser

From irefindex
Revision as of 15:34, 9 July 2010 by PaulBoddie (talk | contribs) (Attempted to make a page corresponding to the revised MITAB format.)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

A tool has been developed to parse the MITAB files produced in the iRefIndex Build Process. Currently, the tool is capable of parsing the MITAB format described on the page README_iRefIndex_MITAB2.6_7.0.

Obtaining the MITAB Parser

The parser and associated resources can be obtained from this location:

Using CVS with the appropriate CVSROOT setting, run the following command:

cvs co mitab

The CVSROOT environment variable should be set to the following for this to work:

export CVSROOT=:ext:<username>

(The <username> should be replaced with your actual username.)


The following programs are required to use the parser:

Running the Parser

Given a directory for the iRefIndex output files such as...

/home/irefindex/output the parser as follows:

python /home/irefindex/output/All.mitab.03042009.txt

It will be necessary to change the date details included in the above filename to match the actual name of the appropriate file found in your own output directory.

Creating the Database

A database can be created using the usual PostgreSQL tools:

createdb -E unicode mitab_irefindex

This database is initialised as follows:

psql -f init_mitab.sql mitab_irefindex

Should the database tables need to be dropped (perhaps in case of problems with the import), the following command can be used:

psql -f drop_mitab.sql mitab_irefindex

Populating the Database

The database is populated as follows:

python mitab_irefindex

As a result, a number of tables representing the structure of the data should be available in the database. For applications built to use this data, indexes may need creating in order to make querying more efficient.

Notes on the Populated Database

The schema used by the populated database attempts to model the data as effectively as possible using a number of tables:

Entity type Tables Table purpose Notable columns Source columns (if different or converted)
Interaction mitab_interactions Model each interaction referencing interactors rigid, intType, edgetype, numParticipants, crigid
mitab_sources Represent sources for each interaction rigid, sourcedb, name sourcedb
mitab_interaction_type_names Represent interaction types for each interaction rigid, code, name interactionType
mitab_interaction_identifiers Represent interaction identifiers for each interaction rigid, dbname, uid interactionIdentifiers
mitab_confidence Represent confidence scores for each interaction rigid, type, confidence confidence
Experiment mitab_method_names Represent detection methods for each interaction rigid, code, name method
mitab_authors Represent publication authors for each interaction rigid, author author
mitab_pubmed Represent publication identifiers for each interaction rigid, pmid pmids
Interactor mitab_interactions Model each interaction referencing interactors uidA, uidB, intType, taxA, taxB, atype, btype
mitab_aliases Represent aliases for each interactor uid, intType, dbname, alias uidA or uidB, aliasA or aliasB
mitab_alternatives Represent alternative identifiers for each interactor uid, intType, dbname, alt uidA or uidB, altA or altB
mitab_interactor_rogs Represent alternative integer identifiers for each interactor uid, intType, rog uidA or uidB, irogA or irogB

Some changes in representation occur when creating the database:

  • Prefixed values are generally split to expose the prefix and identifier, name or value following it.
    • The various interaction and interactor prefixes (such as irefindex:, rigid:, rogid:, crigid: and crogid:) are omitted from interaction and interactor columns.
    • Source identifiers are split with the prefix (such as intact:) used to make a dbname column with the actual identifier stored in its own column (such as alias or alt).
  • The "empty value" (-) should never appear as an identifier, and where such a value is used in a list, that element should be excluded. This is pertinent in the case of vocabulary terms where MI:0000 might be used together with an empty list of identifiers or names as an "empty collection" indicator.
  • Duplicate values in lists are generally discarded.

Further work may include the introduction of a separate interactor table, collecting related information for each interactor.

Canonical interactors and interactions

An intType column has been introduced into some tables in order to indicate whether an interactor or interaction involves canonical information.

  • Where an interactor is a specific, observed interactor, intType will be set to S
  • Where an interactor is a canonical group, intType will be set to C
  • Where an interaction involves only specific, observed interactors, intType will be set to S, and the crigid column will refer to the rigid column of the associated canonical interaction
  • Where an interaction involves canonical groups, intType will be set to C

Thus, the mitab_interactions table effectively has two levels:

  • A "parent" level describing interactions between canonical groups, grouping together records in...
  • A "child" level describing interactions between specific, observed interactors, each referencing a parent record

All iRefIndex Pages

Follow this link for a listing of all iRefIndex related pages (archived and current).