A flexible graph-based controlled vocabulary engine

1
A flexible graph-basedcontrolled vocabulary
engine

• Johann Visagie
• ltjohann_at_egenetics.comgt

2
Background

• Implementation of a controlled vocabulary engine
• Basis for a more complex profiling system that
  will aid in the identification of disease gene
  candidates by integrating
• transcript information
• standardised controlled vocabulary of expression
  terms
• genomic sequence
• genetic mapping information

3
Structure of the Vocabulary

• Orthogonal set of hierarchical schemas (trees)
• Each schema describes an expression domain, e.g.
• Anatomical site, Pathology, Development Stage,
  Cell Type
• A tree's nodes are associated with terms
  describing expression states in that tree's
  domain
• Mapped 6937 cDNA libraries (incl. dbEST, SAGE),
  each with one or more nodes in as many trees as
  possible

4

5
Graph-based implementation

• 2nd iteration
• Python modules implementing hierarchical data
  structures, based on generalised graph library
• Flexible enough for future experimentation
  (different data structures, multiple relationship
  types, etc.)
• All operations in-memory
• Overcomes most limitations of prior
  implementation
• Forced unique terms, limited to pure trees, speed
  issues, database-centricity

6
Query language

• Parser for a simplistic Boolean query language
• pathologycancer AND (anatomyliver OR
  anatomystomach)
• Implicit "query sets"
• Tool for the power user
• Each query term resolves to set of nodes in a
  tree (the node matching the term, and all its
  children), which maps to set of cDNA libraries
• Note Multiple orthogonal classification domains
  allow for construction of almost arbitrary query
  resolution

7
Interfaces

• Python API
• Under development
• SOAP v1.1
• DAS v1.5 (under investigation)
• wxPython-based GUI
• Curation
• Query interface for users

8
Application

• Proved its worth in a number of SANBI research
  projects
• Components of controlled vocabulary system are in
  use by a number of groups

9
Acknowledgements

• Soraya Bardien-Kruger
• Alan Christoffels
• Tania Hide
• Winston Hide
• Paul Hüsler
• Janet Kelso
• Damian Smedley