RProteomics Interoperability Review

1
RProteomicsInteroperability Review

• Patrick McConnell1,Salvatore Mungal1, Richard
  Haney1, Mark Peedin1
• 1Duke Comprehensive Cancer Center

2
RProteomics Team

• Duke, ICR Developer
• Patrick McConnell, Project lead
• Richard Haney, Architect and developer of
  statistical systems
• Salvatore Mungal, Middle-tier Java developer
• Mark Peedin, Database developer
• Northwestern University, Collaborator
• Simon Lin, Proteomics domain expert
• Oregon Health Sciences University, ICR Adopter
• Shannon McWeeney
• Veena Rajaraman
• University of Pennsylvania, ICR Adopter
• David Fenstermacher
• Craig Street
• University of North Carolina, Collaborator
• Cristoph Borchers

3
RProteomics Overview

• Statistical routines to analyze proteomics data
• MS and LC-MS data
• Integrative Cancer Research Workspace
• Proteomics Special Interest Group
• Adopters
• University of Pennsylvania
• Oregon Health Sciences University
• Architecture Workspace
• Reference implementation for analysis services
• Project ends October 1, 2005

4
RProteomics Focus

• We are NOT concerned with
• LIMS and database management
• Identification of proteins
• Database searching and pattern matching
• Statistical Modeling of Spectra (SMOS)
• Data standards
• Generic statistical data and spectral data
• Analysis
• Spectra processing and analysis audit trail

5
RProteomics Details

• Statistical routines to analyze MS and LC-MS data
• Background removal, denoising, alignment/calibrati
  on, normalization, peak finding, isotope
  deconvolution, peptide quantitation, high-level
  modeling
• Open Statistical Services (OSS)
• Bridging Java and R with web services and XML
• Proteomics database
• Data model (XML Schema), object model (Java), XML
  database (Mako)
• Grid services
• Data access, data transformation, and analysis
• Graphical user interface
• Load and query data, run analytics, view plots
• Future plans
• Multilevel and hierarchical statistics
• Dynamic GUI
• Peptide/protein identification services
• Integrate with PIR, Q5, caMassClass, proteomics
  repository

6
Statistics Denoising
7
Modeling Process
XMI Class Attribute Association Definition As
sociation multiplicity
XML Schema Complex types Elements with in-lined
complex types Elements of complex
types Elements of simple types Attributes Anno
tations minOccurs/maxOccurs

• Mapping Rules
• Remove Type from end
• Map schema types
• Add id attitribute
• Move single-attribute classes to an attribute

8
Data Model Overview

• mzXML
• De facto standard for encoding raw proteomics
  data
• ScanFeatures
• Generically encodes proteomics data and
  analytical results
• AML-routine
• Describes analysis routines in great detail
• Metadata to help researchers understand grid
  services
• AML-run
• Keeps track analysis routines (provenance)
• Hooks input data and output data together
• StatML
• Generic encoding for statistical data (lists,
  arrays, and scalars)
• Service parameters
• Parameters to the operations of the grid service

9
mzXML

• Encodes raw spectra data (mz-intensity pairs)
• De facto standard by Sashimi
• Instrumentation
• Data processing
• Separation technique
• Spot description
• m/z scan values
• MALDI acquisition
• Data integrity
• Other candidates mzData and mqData

http//sashimi.sourceforge.net/software_glossolali
a.html
10
ScanFeatures

• Support for statistical data
• Metadata
• project, patient, fraction, replicate, date,
  outcome, scanNumber, scanStartPos, scanStepSize
• Features
• Name
• m/z, intensity, peakWidth, peakHeight, etc.
• Controlled vocabulary
• Value
• Scalar or array
• Support for a hierarchy of features

11
AML-routine

• Authors
• Writers of the routine and contact information
• Routine name
• One-word description
• Title
• Formal one-sentence description
• Aliases
• Alternate names
• Ontological description
• Controlled vocabulary (CDEs)
• Textual description
• Human-readable description (1-3 sentences)
• External references
• Journals, websites
• Pseudo-code
• Approximation of actual code

• Source code
• Actual source code
• Routine Signature
• Input/output parameters
• Contract
• Pre and post conditions
• Usage
• Textual description and examples of how the
  routine is to be used
• Implementation
• OS/hardware/compiler on which the routine is
  implemented
• Caveats
• Any user comments not previously covered
• Benchmarks
• Theoretical performance and links to performance
  runs

12
AML-run
AML-run
lsid
AML-routine
inputs
outputs
lsid
lsid
data
data
lsid
lsid
data
data
lsid
lsid
data
data
user submit time complete time
13
AML-run Cont.
data
data
data
data
aml-run
aml-run
Input can be used for more than one run
Different levels combined together
data
data
aml-run
aml-run
Output can be input for another run
Track back the final output to find what analysis
was performed and what data was used
data
14
StatML

• Scalars
• String, boolean, integer, long, float, double
• Arrays
• Base64 encoded (turn 3 bytes into 4)
• Square arrays (multi-dimensional)
• Integer, long, float, double
• Lists
• Can contain lists, arrays, scalars
• Null
• Lack of value

Efficient, textual encoding
15
(No Transcript)