RAD to MAGE via ObjectMap

1
RAD to MAGEvia ObjectMap
Angel Pizarro

• Platform independent rules for object
  transformation.
• Well, more like semi-dependent, but whos gonna
  read the code?

2
Background

• RNA Abundance Database (RAD)
• Part of the Genomics Unified Database (GUS)
  relational database system
• http//www.gusdb.org, http//www.cbil.upenn.edu
• MAGE-ML 1.0 is the accepted standard for gene
  expression data exchange.
• ArrayExpress, a public repository for gene
  expression data, will only accept MAGE-ML

3
GUS Supports Multiple Projects
AllGenes
PlasmoDB
EPConDB
Java Servlets
Other sites, Other projects, e.g. GeneDB

Oracle RDBMS
Object Layer for Data Loading
Available at http//www.gusdb.org
4
Simple mappings

• One row One object
• CompositeElement ? CompositeSequence
• One Row Multiple Objects
• Element ? Feature (location)
• Element ? FeatureInformation (?mismatch?)
• Element ? FeatureReporterMap (link ?)
• Element ? Reporter (sequence info)

5
Complex mappings

• Array, Assay ? Array, ArrayGroup,
  ArrayManufacture, ArrayDesign, PhysicalArrayDesign
  , Zone, ZoneGroup, ZoneLayout
• Map of subclasses is not uniform.
• Contact ? Person, Organization
• BioAssay ? PhysicalBioAssay, MeasuredBioAssay,
  DerivedBioAssay

6
Our Approach Design decisions

• RAD Perl object-relational layer to MAGEstk Perl
  Object layer
• Try for platform independence for portability
• Looking forward to a RAD java object layer
• Use special purpose code to reduce time to
  market

7
ObjectMap Framework

• ObjectMap collection of RuleSets
• RuleSet collection of Rules
• Rule transform a set of inputs to new set of
  objects
• DataSource the inputs
• ClassRule the outputs
• SlotRule the attributes or associations of a
  class
• Function interface definition for contained
  functions
• Val generic (name, type, value) tuple used to
  set values or differing types

8
The Platform Dependent Engine
Script
RAD object
MAGE object(s)
RuleSet
Functions.pm
ObjectMapEngine

• The rule structure is platform independent (XML)
• The engine that uses the rule definitions is not.
  Currently in Perl, but looking at Java for GUI
  development.

9
Export Example
10
ZoneGroup Rule

• ltDataSource name'array' /gt
• ltClassRule name'ZoneGroup'
• className'BioMAGEArrayDesignZo
  neGroup'gt
• ltSlotRule name'zonesPerX' type'attr'
  method'setZonesPerX'gt
• ltVal type'data' value'array.getNumGridRo
  ws/gt
• lt/SlotRulegt
• ltSlotRule name'zonesPerY' type'attr'
  method'setZonesPerY'gt
• ltVal type'data value'array.getNumGridCo
  lumns /gt
• lt/SlotRulegt
• ltSlotRule name'zoneLayout' type'assoc
• method'setZoneLayout'gt
• ltVal type'rule' value'Export.Array.ZoneL
  ayout' gt
• ltVal name'array' type'data
  value'array /gt
• lt/Valgt
• lt/SlotRulegt
• lt/ClassRulegt

11
Export Example
12
Alternative approaches

• Thought about a code_block Function
• Imbed code directly into the XML
• Makes a mapping platform dependent
• Do it in the application
• Easier data manipulation
• Recursive rule calls can be avoided when too
  complex.

13
What actually happened

• Now do the ZoneGroups and Zones
• my arrows A-gtgetNumArrayRows()
• my arcols A-gtgetNumArrayColumns()
• my gridrows A-gtgetNumGridRows()
• my gridcols A-gtgetNumGridColumns()
• for (my ac 1 ac lt arcols ac)
• for (my ar 1 ar lt arcols ar)
• create a ZoneGroup ZoneLayout
• my zg ENGINE-gtmakeObject('Export.Array.Z
  oneGroup', 'array' gt A)
• for (my gr 1 gr lt gridrows gr)
• for (my gc 1 gc lt gridcols
  gc)
• create the Zones
• my z ENGINE-gtmakeObject('Export.Arra
  y.Zone', 'array' gt A)
• my zone_key "ac.ar.gc.gr"
• z-gtsetIdentifier(z-gtgetIdentifier() .
  "" . zone_key)
• z-gtsetName(z-gtgetName() . "" .
  zone_key)
• z-gtsetRow(gr)
• z-gtsetColumn(gc)
• zg-gtaddZoneLocations(z)

14
Import Example
?
15
Import Example
16
Future Direction

• Refine the rule structure
• Create a Rule Dictionary
• List inputs needed, outputs produced, and the
  purpose of the rule
• Make a Java version
• Allow GUI for the dictionary
• Drag and drop of one object to create another,
  ala Clio project
• http//www.almaden.ibm.com/software/km/clio/index.
  shtml
• IBM Research Clio at google.com.

17
Thanks
http//www.cbil.upenn.edu

• Chris Stoeckert
• Elisabetta Manduchi
• Greg Grant
• Hongxian He
• Junmin Lui
• Trish Whetzel
• Matt Mailman
• Steve Fischer
• Jonathan Schug
• Jonathan Crabtree

http//www.cbil.upenn.edu/angel/code