caGrid Version 0.5 Reference Implementation caArray caBIG Architecture Workspace Face to Face Georgetown University August 16th -18th, 2005

1
caGrid Version 0.5 Reference ImplementationcaArra
ycaBIG Architecture Workspace Face to
FaceGeorgetown UniversityAugust 16th -18th,
2005

• Colin Freas
• Lombardi Comprehensive Cancer Center
• cef6_at_georgetown.edu

2
Outline

• High Level Overview of caArray
• Data Model
• Project Architecture
• Process of getting to Silver level compliance
• Functionality Exposed to Grid
• Process of Grid Enablement
• Demo/Screenshots
• Lessons Learned / Technical Difficulties / Wish
  List
• Acknowledgements

3
Project Overview

• In our context, NCICB's caArray consists of a
  microarray database and the associated web-portal
  and API for accessing that data.
• caArray is a standards based data repository of
  microarray experiment data.
• The MIAME standard describes the Minimum
  Information About a Microarray Experiment that is
  needed to enable the interpretation of the
  experiment results unambiguously and,
  potentially, to reproduce the experiment.
  caArray is compliant with the MIAME 1.1 standard.
• Concepts within the MIAME standard can be mapped
  to the MAGE-OM and to MAGE-ML which are an object
  model and an XML format, respectively. This is
  the basis for caArrays object model.
• caArray consists of two parts a web application
  where users input experiment data and perform
  searches, and the MAGE-OM server, which allows
  programmatic access to caArray objects. API
  access is read-only and is achieved via RMI.

4
Data Model

• Experiment - The Experiment is the collection of
  all the BioAssays that are related by the
  ExperimentDesign.
• ArrayDesign - Describes the design of an gene
  expression layout. In some cases this might be
  virtual and, for instance, represent the output
  from analysis software at the composite level
  without reporters or features.
• BioDataCube The experiment data itself.
• The rest of the is ostensibly required to
  duplicate experiment results. This has been an
  ongoing challenge for microarray experimenters,
  which the MAGE-OM is an effort to address.

5
Project Architecture
6
Process of getting to Silver level compliance

• Programming and Messaging Interface
• RMI based MAGE-OM API and EJB APIs provide
  programmatic access to data.
• Vocabularies/Terminologies Ontologies
• Utilizes MGED Ontologies and corresponding
  vocabularies.
• Data Elements
• CDEs defined and entered in caDSR. caCoreToolkit
  used to define and semantically map CDEs in
  caDSR.
• Information Models
• UML class diagram defined and available for
  interface.

7
Functionality Exposed to the Grid

• caArray is a data service.
• All objects within the MAGE-OM, which are exposed
  via an API are accessible. Important examples
  BioDataCube, Experiment, ArrayDesign.
• Login and password are required, however public
  data from caArray nodes can always be accessed.

8
Process of Grid Enablement

• Standing up caArray locally was no easy task.
  Tremendous help was provided by NCICB to work out
  bugs in the 1.2 release packages, and by Scott Li
  and John Osborne in standing up the MAGE-OM
  server.
• Standing up local caGrid node easier. Scripts
  provided by NCICB installed GLOBUS, OGSADAI,
  Tomcat, the caGrid infrastructure, and the
  caArray node.
• The MAGE-OM server is the interface between the
  grid and caArray.
• Much work in enabling caArray to function on the
  grid had already been accomplished.

9
Demo and/or Screenshots

• Lets go to the videotape!
• Query to caArray public data from Georgetown and
  NCICB instances. Private data from Georgetown.
  Use Experiment objects.
• Query to caArray get gene accession numbers for
  a set of genes in the ArrayDesign for the
  Affymetrix 133 2.0 microarray. Run query with
  accession numbers against caBIO.
• Query to caArray Genepix ArrayDesign for IMAGE
  Clone annotations. Use returned results to query
  caBIO with appropriate Clone characteristics.
• Query to caArray data at Georgetown and NCICB
  returning an aggregate result set.

10
Lessons Learned / Technical Difficulties / Wish
List

• caArray QC. Version 1.2 had a lot of problems.
  It was not possible to install without additional
  resources, provided in our case by NCICB and
  fellow adopters. It took a month to correct the
  files on the public download site. Development
  cycle may be too short.
• caGrid software made the installation process
  very easy, but not necessarily transparent. More
  documentation and copious feedback during the
  install process would be useful.
• Firewall issues. Several ports need to be
  correctly configured on both the MAGE-OM server
  and on the local grid node. A standard way of
  documenting such requirements for caBIG projects
  would be nice.
• Some way to stress test the system and to collect
  performance metrics would be useful.

11
Acknowledgements

• Arnie Miles Georgetown caBIG lead
• Nick Marcou coordinating with caGrid team to
  install node.
• Jack Yuelin Zhu entering test data and testing
  caArray from a users perspective
• Colin Freas overall coordination from
  Georgetown, setting up local caArray instance
• Sumeet Muju Juergen Lorenz indispensable
  support in standing up our local instance of
  caArray
• Andrew Shinohara provided great help in
  elucidating test scenarios and writing the XML
  for the grid queries
• Ruowei Wu Manav Kher providing some excellent
  scripts for installation of caGrid node
• William Sanchez caGrid team liaison,
  coordinating efforts from the NCICB and keeping
  the efforts moving forward with a sense of
  urgency
• Scott Li and John Osborne Scott wrote a
  critical script which correctly ran the MAGE-OM
  server which caGrid communicates with, and help
  in running unit tests against it. John provided
  several example configuration files, and some
  other critical unit tests.