Function Genomics Experiment FuGE Object Model in Plain English

1
Function Genomics Experiment (FuGE) Object Model
in Plain English

• Angel Pizarro
• Institute for Translational Medicine and
  Therapeutics
• University of Pennsylvania
• Andrew Jones
• Department of Computer Science,
• University of Manchester

2
Overview

• Important background information
• Major FuGE concepts, in plain English
• This is just a preamble to the real work
  presented by Andy Jones

3
Vitally Important Concepts

• UML/Object Models ! data standards
• Workflows, from a CS perspective
• FuGE Concepts
• Identifiable
• Ontology
• Protocol
• Material and Data Workflow
• FuGE is more a framework for developing standards
• FuGE is a work-in-progress
• Community feedback needed and encouraged!

4
UML gt data standard

• Data standards are contracts between source and
  consumer of certain types of information
• Range in complexity and details
• Object models are for software
• They sometimes include serialization, which has
  been seen as the data standard
• They have a whole host of requirements that are
  not related in the least bit to the data standard
• Domain Experts should concentrate on Use Cases
• Software Developers of the model must satisfy
  those Use Cases to the satisfaction of the Domain
  Expert, in language they can understand

5
Working directly with object models not
recommended for average user
6
Workflow

• Procedures in the lab, like 2D gel image analysis
• Workflow Management Systems (WFMS)
• Business Process Management
• Several standards for encoding workflows
• XPDL, BPEL, jPDL

7
Example workflow (XPDL)Submit peptide search to
multiple search algorithms

• Resources swim lanes
• States Activities boxes
• Transitions arrows

8
WFMS are cropping up all over the place

• Business Processes
• Data analysis pipelines
• Clinical Trial management
• Laboratory Information Management Systems (LIMS)

9
LIMS Example
Matrix Integra
10
What was the point again?

• Simple and flexible data models do work
• Just Resources, States, and Transitions
• Using them with ontological constraints works
• Part of the State, as pre- and postconditions
• Constraints are specific to the system, but can
  still be shared
• Does not exist in a vacuum
• Layers of standards that work together
• LDAP, WSDL, XML Schema
• The concept of workflows are important to FG

11
End of background

• We now begin to explain the conepts necessary to
  understand FuGE as a whole.

12
FuGE in almost plain english

• The Identifiable type in FuGE uses LSID
  specification for unique identifiers for objects
• LSID Life Science Identifier
• Basically a name-space qualified identification
  string
• Allow us to reference objects not in the document
• Seems to come up as an issue a lot in PSI

13
FuGE OntologyTerm

• A reference to an outside source of terms
• Represent compound terms using a simple nesting
  mechanism
• Usage
• Provide the source of the term
• Provide the accession of the term
• No accession, then no reference needed
• Can be used as a simple controlled vocab
• No ontology inference should be done from a
  reference model!!!
• E.g. you should never evaluate a partial ontology
  statement outside the system it is defined in.

14
FuGE Protocol

• Protocols have a set of ordered simple (atomic)
  actions
• Protocols can have associated Software and/or
  Equipment
• Protocols, Software and Equipment can have a set
  of defined Parameters
• Parameters can have a default value
• Protocols can be structured in a hierarchy to
  define a whole procedure

15
To use FuGE Protocol

• You must first define the Protocol
• Then you describe what actually occurred when you
  applied the Protocol
• Protocol
• ProtocolApplication

16
FuGE Workflow
Material
Material
Treatment
Treatment
Material
Material
Treatment
Data Acquisition
Data Transformation
Material
Data
Data
17
FuGE Protocol Types

• Take physical inputs to produce physical outputs
• Take physical inputs and produce data
• Take data as input and produce new data
• MaterialTreatment
• DataAcquistion
• DataTransformation

18
Usage Scenarios

• Use as-is and provide an ontology to define
  constraints between the states, ala XPDL
• Build on top of the model itself and subclass
  components you feel could use additional typing
  or constraints
• Create your own model, providing some way to
  reference FuGE instances
• A mix of the above
• Not at all

See the presentation by Andy Jones for an
example of this
19
Thanks

• MGED
• PSI
• http//fuge.sourceforge.net