Gene Expression RFP Initial Submission

1
Gene Expression RFP Initial Submission

• Scott Markel, Carl Foeller
• NetGenics, Inc.
• 11 December 2000

2
Overview

• Background
• CORBA/OMG approach
• Domain details
• General approach
• Model objects with examples
• Types of queries to support
• Request for feedback
• Acknowledgements

3
Background

• RFP issued on 10 Mar 00 (lifesci/00-03-09)
• Initial submission
• document lifesci/00-11-06
• XMI lifesci/00-11-07
• IDL lifesci/00-11-08

4
Modeling Approach

• UML model is normative
• UML permits semantic specifications that go
  beyond what is expressible in IDL
• UML follows UML Profile for CORBA
• XMI and IDL representations generated by Rational
  Rose
• Aligns with new Model Driven Architecture

5
Abstract Interfaces and Valuetypes

• Both abstract interfaces and valuetypes are used
  to represent the largely data centric objects in
  the model
• Valuetypes support the abstract interfaces and
  contain private data members
• The use of abstract interfaces provides for
  greater flexibility than valuetypes alone provide

6
Abstract Interfaces and Valuetypes (contd)
7
General Approach

• View from perspective of science(mirror the
  physical)
• Objects reflect
• Physical objects and properties (e.g., arrays)
• Operations and relationships between objects
  (e.g., spot and array)
• Data and analytical concepts

8
General Approach (contd)

• Data Separation
• Platform from measurement
• Objects have inherent properties independent of
  experiment
• Observation from the analysis
• Record layer of raw data and layer of
  normalized/cleaned/analyzed data
• Maintain technology independence and aim for
  general extensibility

9
Experiment Overview

• Gives overview of main objects and relationships
• Distinguishes replicate experiments from repeated
  measures
• Shows how the probe can have multiple sample
  sources and labels

10
Experiment Overview (contd)
11
Object Overview
12
Model Objects

• Source - biological origin of probe, spots
• Probe - mobile phase, labeled sample
• Spot - oligos or cDNA to be probed
• Array - physical support for spots
• Experiment - assay of probe vs. array spots
• Project - collection of experiments

13
Source

• Organism
• Tissue
• CellType
• Gender
• Disease
• Developmental stage

• Age
• Genotype
• Phenotype
• Supplier
• Isolation method

14
Probe

• Material in mobile phase of Experiment
• ID
• Source (1 or more - can be mixture)
• Type of label (1 or more - can be mixture)

15
Spot

• Spot contains non-experimental information about
  the physical spot
• Gene name
• Sequence
• Parent sequence (with database Identifier)
• Location on array (row, column)

16
Array

• Array contains non-experimental information
• ID, manufacturer, model, type (e.g., glass)
• Spots and background spot
• Source
• Spotting method (e.g., inkjet)
• SpotType (e.g., oligo)
• Dimensions (rows, columns)

17
Experimental Data

• SpotValue - expression level and normalized
  expression level (two floats per channel)
• SpotData - SpotValues vector and fold change
  allows multiple channels
• ArrayData - 2-D array of SpotData
• SpotData SpotValue extended for specific
  technologies (Affymetrix and Incyte)

18
SpotValue Extensions
19
SpotData Extensions
20
Experiment

• Name
• Probe ID
• of arrays, array Ids, ArrayData
• ExperimentType (e.g., time series)
• Sample Treatment (i.e., exptl protocol)
• Probe isolation method

21
Experiment (contd)

• Hybridization treatment(s)
• Measurement Type (e.g., Cy3) must equal Probes
  label
• Control
• Researcher, organization, date
• Normalization Information
• Related experiments

22
Project

• A collection of experiments (ExperimentCollection)
  
• ExperimentCollectionType (e.g., control)
• get/set_experiments()
• add_experiment()
• contains_experiment()
• Contains related experiments (ExperimentCollection
  s)

23
Treatments

• Sample Treatment
• Experimental protocol - dosage, time series or
  compound comparison
• including details like heat-shock, starvation
• Plain, Radiation, Compound, Other
• Hybridization Treatment

24
Treatments Extensions
25
Controls
26
Replicate Set

• Collection of Experiments
• Members have same
• ArrayType
• Spots
• Treatments
• Probe isolation method

27
Related Issues

• Clusters - sequence or expression profile
  similarity
• Related spots can be indexed and addressed as
  desired
• Related spots can be identified as result of
  queries
• Can base new query on results of previous (e.g.,
  do genes sharing expression profile have links to
  similar KEGG functional pathways?)
• Additional clustering falls in the category of
  gene analysis (new RFP?)

28
Related Issues (contd)

• Annotations
• Can be added to sets of spots as result of query
• Multiple array experiments
• Are treated like a single virtual array
• Spots independently addressable
• Analyses and statistics can be taken over whole
  set

29
Querying

• GeneExpressionQueryEvaluator extends
  CosQueryQueryEvaluator
• Query language details yet to be determined

30
Example Queries

• Show experiments that exist in project X
  performed by researcher Y
• Show all experiments performed on mice, in any of
  the three tissues A, B, or C, where the fold
  change for gene X is greater than 2.0
• Retrieve all genes in the time series experiment
  ABC that were differentially regulated across the
  time series

31
Request for Feedback

• Queries, query language (SQL)
• Externalization (GEML, MAML, other XML DTDs)
• Support for image needed?
• Other comments

32
Acknowledgements

• Michael Dickson
• David George
• Ken Griffiths
• Ty Jacobs
• Chris Sears
• Jan Weaver