1. Clinical Proteomic Tumor Analysis Consortium: Ontology Considerations
Interrogating Cancer Biology to Address Clinically Relevant Questions
Clinical Proteomic Tumor Analysis Consortium: Ontology Considerations
Effect of post-excision-delay-to-freezing time on posttranslational modifications of tumors: a study by the NCI Clinical Proteomics Tumor Analysis Consortium (CPTAC)
Authors: Philipp Mertins, D. R. Mani, Tao Liu, Feng Yang, Aaron Gajadhar, Hannah Johnson, Hui Zhang, Douglas A. Levine, Reid Townsend, Sherri Davies, Michael Gillette, Vladislav Petyuk, Karl Clauser, Jana Qiao, Jeffrey Whiteaker, Gordon Mills, Ronald Moore, Narciso Olvera, Fanny Dao, Daniel Chan, Daniel Liebler, Karin Rodland, Richard D. Smith, Amanda Paulovich, Matthew Ellis, Forest White and Steven A. Carr, NCI-CPTAC Consortium
Abstract: 300/300 word limit
The NCI-CPTAC Consortium is characterizing the proteomes of large numbers of breast, colon and ovarian tumor samples that have been genomically characterized by The Cancer Genome Atlas (TCGA) program of the NCI. In addition to global proteome analysis, we also plan to study posttranslational modifications in these samples, a unique contribution that proteomics can make. The time between tumor excision and freezing (Post-excision delay time or PDT) for the TCGA samples varied from minutes to over an hour.  Evidence from reverse phase protein array studies employing phospho-specific antibodies showed that phosphorylation stoichiometry of specific proteins can change within minutes due to the trauma of excision, hypoxia and ischemia.  However, only a very small subset of the phosphoproteome could be analyzed in these studies, and other modifications of interest such as glycosylation or acetylation were not analyzed. In addition, time points earlier than ca. 15 minutes were not studied.
Before embarking on analyses of PTM's in these samples, we wanted to understand the potential effects of PDT. Four time point samples (0, 5, 30, 60 min PDT) of xenografted human breast cancer tumors and patient-derived ovarian cancer tumors were created by excising tumor samples prior to vascular ligation so that ischemia time was accurately defined. Samples were analyzed for global, quantitative proteome, phosphoproteome and N-linked glycome using iTRAQ quantification and high performance, multidimensional LC-MS/MS on Orbitrap mass spectrometers. While no significant changes in the proteome were detected across the timepoints, and the vast majority of phosphosites were also stable, changes at specific phosphosites were observed as early as 1 minute post-excision. The biological processes affected involved stress, cell death and various kinase activation pathways.  Both kinase and phosphatase activation as a result of PDT were evident. The results of our phosphopeptide as well as other PTM analyses will be described.
Presented by Chris Kinsinger
NCI/CSSI/OCPPR
May 12, 2015

2. TCGA tumor collections
Map proteome/PTMs to each patient’s genome; develop assays for pathways and candidate biomarkers
Inputs
Analyses
Outputs
TCGA tumor collections
Analyze >100 tumors/cancer
breast, ovarian, colon
Proteome
Phosphoproteome
Other PTMs
Targeted protein / PTM
Cancer pathways
Protein isoforms
Molecular signatures
Genome-proteome relationships
Genome-signaling relationships
Protein targets
Assays for newly discovered proteins
Prospective tumor
collections
NOTE: 100 – 150 tumors per cancer to be analyzed
Biological mechanisms:
Are genomic aberrations detectable at protein level?
What is their effect on protein function?
Which events are drivers? Which are passengers?
Clinical applications:
Can proteomic information provide a better molecular taxonomy of cancer?
Can genotypic information guide protein marker development?
The Centers: Broad/FHCRC; PNNL; Vanderbilt; Wash U.; Johns Hopkins

3. TCGA CrCa (transcriptomic molecular subtypes)
Identified three transcriptomic subtypes (mRNA clusters): MSI/CIMP, Invasive, and CIN
MSI/CIMP subtype is enriched with hypermutated tumors
Question: Can proteomics data rediscover or redefine colorectal cancer subtypes?
MSI: microsatellite instability
CIMP: CpG island methylation phenotype
CIN: chromosome instability
The Cancer Genome Atlas Network; Nature 487, (2012) doi: /nature11252

4. CrCa: Global proteome reveals 2 new subtypes
Transcriptome Subtypes
MSI/CIMP
Invasive
CIN
genes
Proteome Subtypes A B C D E A B C D E
Hierarchical clustering Heat Map
Highlight B and C vs MSI-CIMP first; note HYP and MSI-hi; note no TP53 or 18q loss; note BRAF and POLE mutations; then show E enriched for TP53 mutations and 18q loss and CIN subtype; then note that B and C split the MSI-CIMP transcriptome classification, but on protein expression; note also that C is associated with Sandanaman ‘stem-like’ and De Sousa CCS3, which are both associated with poor prognosis.
Red – up regulated; Green – down regulated
We started with 5630 proteins and 90 samples. Using the consensus clustering method and the top 1200 proteins with the largest expression variation across the 90 samples, we identified six subtypes. After evaluating cluster significance, one small cluster with four samples was not significant and was removed from further study. Then we used the silhouette plot to select core samples for each subtype, and 12, 10, 12, 28, and 24 samples were identified as core samples for subtypes I-V, respectively (77 samples in total). Using cross-validation, we removed genes that are not essential for subtype definition, and identified signature genes for each subtype (286 in total). The heat map on this slide visualizes protein expression patterns for the 286 genes in the 77 samples. Each row is a gene and each column is a sample. Samples are separated into five subtypes as indicated by the color bars at the top. Hypermutation status is also presented and it is clear that subtype 3 (orange) is enriched with hypermutated samples (pink). On the vertical axis, the color bars indicate signature genes for each subtype (with matching colors). Signature genes can be separated into up or down-regulated ones, relative to other subtypes. For some subtypes, signature genes primarily come from one direction (e.g. Subtype 2 signature genes are mostly down-regulated). For each set of up or down signature genes, enriched GO terms were identified and labeled (e.g. Subtype 4, the purple subtype, is characterized by up-regulation of genes involved in blood vessel development and down-regulation of genes involved in mitotic cell cycle).
De Sousa
patient tumors

5. CPTAC workflow Tissue collection Tissue qualification Data generation
Data analysis

6. Tissue collection Tissue Collection Tissue qualification
Data generation
Data analysis

7. Clinical and biospecimen data
TCGA forms
CPTAC Biospecimen working group
~ caDSR integration

8. Tissue collection Case report forms (CRF) Submission CRF
Basic demographic, clinical, and biospecimen data to ensure case meets inclusion criteria
27 elements
Baseline CRF
Histology
Diagnostic pathology
IHC results (eg ER/PR for breast cancer)
41 elements
1-year follow-up CRF
Status
Surgical margin
Chemotherapy
Additional tumor events
24 elements

9. Additional forms due at 12 months
Other malignancy form
Due if a subject has an additional cancer diagnosis
Diagnostic information
Surgical and treatment data
23 elements
Pharmaceutical form
Chemotherapy
Response to therapy
8 elements
Radiation supplemental treatment form
Radiation therapy
9 elements
Up to 134 data elements

10. Tissue qualification Tissue qualification Tissue collection
Data generation
Data analysis

11. Tissue qualification Criteria No public facing data
Pathology-based criteria
Nucleic acid-based criteria
No public facing data
Challenge: how best to give access to pathology images?

12. Data generation Data generation Tissue collection Tissue qualification
Data analysis

13. Experiment protocol PSI Mass Spectrometry Ontology [MS] (EBI)
Database, Vol. 2013, Article ID bat009, doi: /database/bat009
8 main branches:
Chemical compound
Contact attribute
External reference identifier
File format
Software
Spectrum generation information
Spectrum interpretation
Standard

14. CPTAC Experimental Protocol
Protocol (pdf)
Protocol summary (xlsx)
Analytical Sample Protocol (9 CDE)
Starting amount
Proteolysis
Alkylation
Enrichment
Chromatography Protocol (8 CDE)
Column length
Injected
Inside diameter
Mass Spectrometry Protocol (8 CDE)
Instrument
Resolution
Collision Energy

15. System map of LC-MS performance metrics – ontology for quality?
Over 40 performance metrics monitored
Rudnick et al. Mol. Cell. Proteomics (2010)

16. Data analysis Data analysis Tissue collection Tissue qualification
Data generation
Data analysis

17. CPTAC Common Data Analysis Pipeline (CDAP) at NIST
MS Data Files
Conversion / iTRAQ
Database Search
MS1 Data Processing
Phosphosite Localization
QC Calculations
PSM Report Generation
ReAdW4Mascot2
MSGF+
ProMS
Phospho RS
nist_ metrics
single_file_report
Raw
mzML
mzXML
MGF
MS1
METADATA
MZID
TSV
TXT
XML
MSQC
MSP
PSM
Generalized Parsimony (gene-based)
PEPTIDES.TSV
SUMMARY.TSV
iTRAQ.TSV
SPECTRA_COUNTS.TSV
PRECURSOR_AREA.TSV
PSMLab
MZID
Publicly Distributed at the DCC
N. Edwards (Georgetown)

18. Peptide spectrum match files
PSM
Tab-delimited
22 data elements, including:
MS2 scan number
Peptide sequence
Charge, score, precursor peak
Feeds some higher analysis tools
MZIdentML
HUPO-PSI compliant format
5 elements, including
Scan number
Gene accession number
Element ids for spectra results
Avoids semantic interpretation
Extract PSMs without having to randomly address sequence collections

19. Protein reports: summary.tsv
Goal: identify and quantify proteins in sample based on identified peptides
12 data elements
Gene (NCBI Gene name)
Distinct peptides
Spectral counts
Gene description
Organism
Chromosome
Locus (cytoband)
Proteins (proteins associated with the gene)

20. Challenges Multi-faceted program: how to keep ontology needs updated?
How to improve clinical annotation of biospecimens?
Integrate HUPO-PSI ontologies/CV with cancer-specific resources
Updating ontologies/CV in a rapidly evolving technology field

21. CPTAC Steering Committee & Working Groups
Common Data Analysis Pipeline
Paul Rudnick
Steve Stein
Sandy Markey
Jeri Roth
David Tabb
Sam Payne
NCI
Henry Rodriguez
Emily Boja
Tara Hiltke
Chris Kinsinger
Mehdi Mesri
Robert Rivers
Jerry Lee
Mandie White
Tim Crilley
Leidos Biomedical Inc.
Linda Hannick
Joy Beveridge
Michelle Hester
Kevin Groch
Kathy Terlesky
Ellen Miller
Lenny Smith
Maureen Dyer
Melissa Borucki
Carla Chorley
ESAC/Georgetown
Karen Ketchum
Nathan Edwards
Ratna Thangudu
Peter McGarvey
Shuang Cai
Mauricio Oberti
Biospecimens WG
Molly Brewer
Sherri Davies
Mike Gillette
Doug Levine
David Ransohoff
Steve Skates
Rob Slebos
Mark Watson
David Tarin
Biospecimen Core Resource
Dave Mulvihill
George Bijoy
Melissa McKenna
Brian Goetz
Amy Brink
CPTAC Steering Committee
Steve Carr
Daniel Chan
Xian Chen
Matthew Ellis
Daniel Liebler
Amanda Paulovich
Karin Rodland
Dick Smith
Reid Townsend
Bing Zhang
Hui Zhang
Zhen Zhang

22. Thank you