NCI Proteome Informatics

1
NCI Proteome Informatics

• David J. States, M.D., Ph.D.
• February 8, 2005

2
Genomics vs. Proteomics

• Genome sequence
• One copy of each gene per genome
• Duplicated genes
• Diploid or higher ploidy
• No tissue variation
• Somatic cell recombination
• Modification not an issue
• Methylation
• Few sample handling issues
• DNA is DNA is DNA
• Very stable

• Proteome
• Wide range of expression levels
• 10-12 orders of magnitude
• Quantification is a goal
• Tissue specific variation
• Plasma proteome
• Post translational processing
• Many known modifications
• Potentially novel chemistry
• Sample processing matters
• Serum vs. plasma
• Many protocols

3
Errors and UncertaintyLessons from Genomics

• GenBank before there were errors
• My lab would never submit an erroneous
  sequence
• Krawetz 1987 survey 1 in 300 nt later revised
  or retracted
• Developing a framework
• Identifying error processes
• Computational representation
• Quantitative error analysis
• Setting standards
• Data use gt accuracy requirements
• Cost vs. accuracy analysis
• Validating lab performance

4
Error Processes in Sequencing

• Genomic DNA
• ?
• Subclone
• ?
• Sangersequencing
• ?
• Sequenceassembly

• Polymorphism
• Degradation
• Sheering, cross linking, oxidation, depurination
• Polymerase fidelity
• cDNA
• Clonal instablity
• Repeats, poison sequences, rearrangements
• Lane tracking
• pre-capillary
• Base calling
• Assembly errors

5
Setting Standards

• Applications drive accuracy requirements
• Homolog identification
• Very error tolerant (evolution as an error
  process)
• Translation of nt to aa sequence
• Frameshift and nonsense errors rare in an ORF
• 1 kb reading frame gt 1/10kb error rate
• Polymorphism identification
• Error rate substantially below the polymorphism
  rate
• Bermuda Quality
• Standard
• No more than one substitution per 10 kb
• No unclosed gaps
• No more than 1 of the sequence in gaps
• Arrived at by community consensus conference
• Does not meet all needs

6
Cost vs. Accuracy in DNA Sequencing
7
Quality Assurance Exercises

• CRADA Funding Mechanism
• Contract with deliverables
• Supervision by NIH staff
• Blind resequencing of test samples
• Until you have done a megabase of sequence, you
  can not really estimate error rates at 1/10kb
• 8 Labs gt 3 Centers
• Major technology choices locked in

8
Proteomics Accuracy Requirements

• Identification
• Reproducibility
• 100 lab to lab reproducibility is not necessary
• Needs to be achievable with reasonable resources
• Accuracy
• Paralog and splice isoform issues
• Biomarkers
• Biological chemistry
• Characterization
• Post translational modifications
• Nature
• location
• Quantification
• Application specific, Issues familiar

9
Abundance vs. Detection
10
Significant But Irreproducible Data

• Identifications may be highly significant even if
  not reproducible
• Sampling issues in the instrumentation
• Biochemical handling
• Proposal Reproducibly within the original lab
• Poisson statistics
• Nobs2 implies ??2
• Another lab reproducing your experiment has a
  high probability of confirming your observation
  if they are willing to put in comparable effort
• P gt 97 at twice your effort

11
Identification Five Levels

• Member of a gene family
• A peptide matching only in this gene family
• Often useful biologically information
• Gene product (transcription unit)
• Multiple peptide matches
• Complete genome sequence available
• Peptide matches that are diagnostic between
  paralogs
• Post translational modification
• Modifications defined by number, type and
  location
• Varying levels of precision (e.g. residue vs.
  peptide level assignment)
• Transcriptional/splice variant
• Multiple peptide matches including matches
  defining the N and C termni
• Peptide matches that are diagnostic between
  paralogs
• Peptide matches or molecular weight data
  diagnostic fo splice isoform
• Complete covalent structure
• Covering set of peptide matches
• Covering set of MS/MS data on all peptides

12
Protein Integration Workflow
.
13
Bacterial Proteins in Normal Blood
Number RefSeq ID Description of
peptides 6 NP_417798.1 Elongation factor EF-Tu
E. coli... 5 NP_415477.1 Outer membrane protein
3a E. coli K12 4 NP_416010.1 Glutamate
decarboxylase isozyme E. coli K12 2 NP_751975.1
Chaperone protein dnaK E. coli
CFT073 2 NP_415333.1 DNA protection during
starvation E. coli K12 2 NP_756310.1 Lipid
A-core, surface polymer ligase E.
coli... 2 NP_418165.1 Low affinity tryptophan
permease E. coli K12 2 NP_416192.1 Murein
lipoprotein E. coli K12 2 NP_418169.1 Hypotheti
cal protein b3713 E. coli K12 2 NP_288226.1 Put
ative AraC-type regulatory protein E. coli...
14
Issues in Project Coordination

• Multiple formats permitted for submissions
• XML
• Pedro data definitions
• Only 2 labs used XML Local informatics
  resources
• Web based submission of Excel templates
• Multiple versions as the project evolved
• Email submission of Excel templates
• Initial suggestion
• Technical concerns, but worked OK
• Some revised submissions Core had to clarify
  whether new data or revised data.

15
Informatics Goals and Mission

• Guide and Coordinate Activities of the Eastern
  Consortium
• Sample tracking between labs
• Data interchange between consortium labs
• Record of project deliverables
• Support beyond the active project period is an
  issue
• Support for project-wide, unanticipated and post
  hoc analysis
• Data interchange and dissemination
• Between the Eastern and Western Consortia
• With the scientific community
• Archival storage???

16
Division of Responsibilities

• Laboratory
• Detailed laboratory records
• Assignment of internal lab identifiers
• Maintenance of association between lab and
  project wide identifiers

• Informatics Core
• Details appropriate for publication
• Assignment of identifiers for
• Samples generating deliverable data
• Samples exchanged between laboratories

17
Informatics Overview
18
Categories of Data

• Three levels of output data need to be provided
  to the Informatics Core for each assay
• Raw data files
• Valuable for reanalysis
• Likely to be instrument vendor specific
• Extracted data files
• Machine and vendor-independent format files
• Describing the data produced by an assay
• tandem mass spectroscopy gt peak lists in .dta or
  .mgf format and search output files
• gene expression analysis gt complete lists of
  gene expression levels
• Analysis results
• The high level conclusions derived from an assay
• mass spectroscopy gt protein identifications with
  the supporting peptide sequence lists and
  associated explicit search criteria and filters
• gene expression gt tables of genes that exhibit
  significant changes in expression with the fold
  change and confidence ranges

19
Choice of caLIMS

• Data definitions based on the NCI caLIMS
  Laboratory Information System (LIMS)
• Many LIMS both commercial and academic
• caLIMS is designed for molecular biology
• caLIMS is an NCI supported project, implemented
  by SAIC and integrated with caBIG
• caLIMS Technical Overview http//calims.nci.nih.go
  v/developers/
• Data definitions provides a common vocabulary
• process vs. protocol vs. project

20
Adapting caLIMS to the MMC

• caLIMS is generic
• No explicit support for proteomics or genetics
• Our focus is on the data definitions and entity
  relationships
• Choice of whether to implement a local LIMS and
  whether to use caLIMS forms and interfaces within
  the lab is entirely up to the lab
• MMC vs. PPP
• Similar in focus on analysis of blood proteins
• MMC will have many more samples and time
  dependent experimental series

21
All Consortium Samples

• Unique identifier
• assigned by the Informatics Core
• Mouse model
• Sample type
• Parent
• Sample or animal from which this specimen was
  derived
• Pooled samples are a special case, details off
  line
• Protocol used to obtain the sample
• Date
• Optional
• see additional fields in the caLIMS SAMPLE table
  below

22
All Consortium Protocols

• Unique identifier
• assigned by the Informatics Core when submission
  received
• Descriptive title
• Text description
• Optional
• see additional fields in the caLIMS PROTOCOL
  table below

23
Animal Models

• Specific model (of several in each lab/each tumor
  type)
• Unique identifier
• Tracking and relationship of samples from the
  same animal
• Background strain
• Genetic manipulations
• Age
• Treatment protocols
• Xenograft information
• Species/cell line
• Site of implant
• Time since implant
• Drugs and exogenous agents
• Agent
• Method and site of administration
• Dosage
• Time since treatment

24
Tissue and Cell Line Samples

• Identifier for the animal of origin
• Treat cell lines as animal lines for informatics
  purposes
• Anatomic origin
• Using GO and Jackson Labs nomenclature
• Protocol used to obtain the sample
• specific protocol identifier and an associated
  text description of the protocol
• Any pooling or aliquoting performed

25
Antibodies

• Reagent identifier
• Species
• Monoclonal vs. Polyclonal
• Immunogen
• If the immunogen was derived from a project
  specimen, the specimen identifier
• Purification or fractionation
• Conjugation or derivatization
• Characterization information
• Specific ligand affinities if known
• Qualitative assessment of suitability for western
  blotting, immunohistochemistry,
  immunoprecipitation
• Cross reactivity if known

26
Proteomics Sample Processing

• Depletion of abundant proteins
• Class
• Agilent column or spin cartridge
• Albumin and IgG only
• IgY antibodies
• Dye or other affinity agents
• Protocol identifier
• Protein fractionation
• Labeling (fluorescent dyes isobaric tags)
• Class
• Electrophoresis vs. chromatography
• Charge vs. size vs. hydrophobicity
• Information about the fraction
• Approximate molecular weight/pI
• Protocol identifier

27
Proteomics Sample Processing

• Cleavage reaction (digestion)
• Enzyme or reaction used
• Specific cleavage protocol identifier
• Pattern describing the cleavage site
• Protocol identifier
• Peptide fractionation
• Class
• Electrophoresis vs. chromatography
• Charge vs. size vs. hydrophobicity
• Information about the fraction
• Approximate molecular weight/pI
• Protocol identifier
• Derivatization
• Chemical reactions performed to prepare the
  sample for analysis
• Blocking groups
• Conversion of lysine to homoarginine
• Sulfydryl modifications
• Other.
• Protocol identifier

28
Mass Spectroscopy

• Sample(s) being analyzed
• Type of experiment (MALDI/MS vs. ESI-MS/MS, etc.)
• Type of instrument used/specific model and any
  modifications
• Protocol identifier
• Data sets
• Raw data files
• Extracted data
• Peaklists
• Search output files (Mascot, Sequest, etc.) with
  explicit criteria/filters
• Analysis
• Protein identifications
• Database(s) used in the search
• Analysis protocol identifier
• Scores for the overall protein identification
• Peptides used to make the identification and
  individual peptide scores
• Quantification
• Analysis protocol identifier
• Relative or absolute units
• Probability estimates for correct and erroneous
  IDs

29
Multi-Dimensional Fractionation

• PF2D, FFE, IPAS, etc.
• Sample(s) analyzed and labeling/derivatization
• Protocol identifier, including pooling of
  fractions
• Data sets
• Raw data file
• Extracted data
• Virtual 2-D gel
• 1-D chromatograms
• Analysis
• Identification
• See MS (above)
• Tracking information needed to related protein
  identifications back to PF2D fractions
• Quantification
• Analysis protocol identifier
• Relative or absolute units
• Error estimates

30
Protein Arrays

• Sample analyzed and labeling/derivatization
• Protocol identifier
• Data sets
• Raw data files
• Array images
• Extracted data
• Spot locations and peak intensities
• Spot assignments
• what was spotted where
• replicate information
• Analysis
• Quantification, including normalization
• Analysis protocol identifier
• Relative or absolute units
• Error estimates

31
Assay Representation

• Project
• Consortium wide
• May involve one or several laboratories
• Protocol
• Description of a procedure
• May be lab specific or used in several labs
• May be used in one or several projects
• Sample
• Specific instance of an experimental sample
• Sample_type table describes classes of samples
• Consortium wide identification
• One sample may be used in one or several labs and
  projects
• Assay
• Specific instance of an experiment

32
Summary

• Proteomics is still in the early days
• Need to better define error processes and
  accuracy requirements
• Informatics support in the labs limited
• Dangers in too rigid an approach to QC
• Project coordination
• Division of labor between labs and the consortium
  data center
• Define deliverables early and inclusively
• Archive of data at multiple levels
• raw, processed, analyzed