1. Proteogenomic Novelty in 105 TCGA Breast Tumors
Karl Clauser
CPTAC Breast Cancer Analysis Group
Broad Institute of MIT and Harvard
Fred Hutchinson Cancer Research Center
Washington University
New York University
CPTAC Data Jamboree
April 16, 2014
National Institutes of Health
Bethesda, Maryland

2. Tumor-specific protein databases for MS/MS-spectra searches
Kelly Ruggles, David Fenyo, NYU

3. QUILTS: Treatment of different variant types
In alternates
frameshifts
Unannotated
Alternative Splicing
1 frame translation
1 frame translation
In frameshifts db
1 frame translation
Novel
Partially
Novel Splicing
Novel
Novel downstream:
1 frame translation
Novel upstream:
6 frame translation
In other db
Completely
Novel Expression
6 frame translation
Fusion Genes
6 frame translation
In variants db
Variants
1 frame translation

4. Proteogenomic mapping: Genetic alterations can be observed on protein level (105 tumors)|
work in
progress
Low confidence thresholds applied to Genome calls
Variants: >2 QUAL phred-scaled quality score in ALT
Alternative splices: >1 read
This document defines the quality value as:
"QUAL phred-scaled quality score for the assertion made in ALT. i.e. give -10log_10 prob(call in ALT is wrong). If ALT is ”.” (no variant) then this is -10log_10 p(variant), and if ALT is not ”.” this is -10log_10 p(no variant). High QUAL scores indicate high confidence calls. Although traditionally people use integer phred scores, this field is permitted to be a floating point to enable higher resolution for low confidence calls if desired. (Numeric)”
Low thresholds applied to Genome calls (>1 read RNA-seq, >2 QUAL phred-scaled Variants)
High thresholds applied to Proteome calls (<0.1% FDR)
% of frameshifts, alternative splices & single AA variants observable by proteomics
mRNA may not be translated or at low abundance
Proteome coverage is incomplete

5. Global proteome and phosphoproteome
discovery workflow for TCGA breast tumors
1 mg total protein per tumor
Internal reference: equal representation of basal, Her2 and Luminal A/B subtypes

6. Serial Search Strategy with Personalized Databases
Concatenated FASTA files, 105 patients
Altered proteins
Removed redundant entries
25,776,160 Spectra
(105 patients)
(36 iTRAQ experiments)
(25 LC-MS/MS runs / experiment)
> Canonical – Variant Patient 1
SIGNALINGPATHWAHREGULATOR
>Canonical Protein – Variant Patient 2
SIKNALINGPATHWAYREGULATOR
Variants: 133,241
3247 Variants Matched
RefSeq-Hs-7/2013: 31,852
> Canonical – Alternate splice Patient 1
SIGNALINGREGULATOR
>Canonical – Alternate splice Patient 2
SIGNALINGPATHREGULATOR
Alternate Spliceforms: 67,853
> Canonical Protein
SIGNALINGPATHWAYREGULATOR
197 Splice Junctions Matched
11,328,955
Matched Spectra
(44% of total)
(1% FDR)
14,447,205
Leftover
Spectra
> Canonical – Truncation Patient 1
SIGNALINGPATFRAMESHIF
>Canonical – Novel Exon Insert Patient 2
SIGNALINGPATHWAYINSERTREGULATOR
>Canonical – Partial Exon Deletion Patient 3
SIGNALINGPATHWAYULATOR
Frameshifts: 19,944
22 Truncation Overlaps Matched
11 Insertion Overlaps Matched
49 Deletion Junctions Matched
Concatenated: 252,890
Low confidence thresholds for Genome calls
Variants: >2 QUAL score (phred-scaled)
Alternative splices, frameshifts: >1 read
High confidence for Proteome IDs
<0.1% FDR peptide spectrum match

7. Frequency of Single AA Variants, Alternative Splices, Frameshifts Across Patients
Somatic variants are less frequent than germline variants
Some germline variants are very common
Rare germline variants present in RefSeq
Some alternative splice forms and frameshifts are very common
Should be in RefSeq
Genome &
Transcriptome
Data
very
common

8. 1 experiment: 3 individual patients + 1 Common control (40 patients)
How many RNA-seq reads to yield a proteomics observation of an alternate splice or frameshift?
1 experiment: 3 individual patients + 1 Common control (40 patients)
197 Alternative splices
82 Frameshifts
Max #
Reads
17 observed
in >1
Expmt
Max #
Reads 19
observed
in >1
Expmt

9. Present in only 1 Common control member
Frameshift Truncation: ras-Related protein Rab-15 Observed only in Proteomics Exp 3
E159
Max RNA-Seq Reads: 1
Present in only 1 Common control member

10. Present in only 1 Common control member
Frameshift Truncation: Cysteine-rich protein 1 Observed in 9 Proteomics Experiments
E159
Max RNA-Seq Reads: 1
Present in only 1 Common control member

11. Present in only 1 Common control member
Frameshift Truncation: Cullin-2 isoform a Observed in 3 Proteomics Experiments
Max RNA-Seq Reads: 1
Present in only 1 Common control member
E159

12. 1 experiment: 3 individual patients + 1 Common control (40 patients)
Many missing observations even when transcript present in many common control members
1 experiment: 3 individual patients + 1 Common control (40 patients)
Alternative splices
Frameshifts

13. Majority of Alternative Splice Junctions and Frameshifts observed in >1 Proteomics Experiment
Pie chart
1 experiment: 3 individual patients + 1 Common control (40 patients)
Alternative splices
Frameshifts
150/197 observed in >1 experiment
44/82 observed in >1 experiment

14. Next steps: Examine “other” category Fusion genes (junction-spanning)
Novel exon splicing (2 sides)
Completely novel gene
Use updated somatic variants from QUILTS
Define genomic data thresholds suitable for proteomic observations
RNA-seq: Min read count
Variant calling: phred-scaled QUAL score
Sort out Germline/Somatic variant call mix status across patients

15. Summary of Proteome Re-processing 105 TCGA patients- 36 iTAQ experiments

16. Changes in Re-processing of TCGA data
Extraction
Centroiding Use Xcalibur , instead of SM.
iTRAQ ratios  are little changed,
intensities lower by ~5x (will more closely match NIST central analysis pipeline)
Precursor  MH+  range expanded from to
Searches
Replace database with RefSeq version used as reference for the personalized database generation.
database content/size very similar,
protein identifiers change from gi numbers to RefSeq numbers.
Allowed modifications will be expanded. Increases the # of identified spectra by ~10%.
From Full iTRAQ, M-ox, N-deam, q-pyro
To iTRAQ-Full-Lys-only, M-ox, N-deam, q-pyro, c-pyro, Ac-nTermProt
Autovalidation
Proteome initial processing, peptide FDR per experiment : %,
but overall peptide FDR across all 36 experiments: ~5.5%
Phosphoproteome initial processing , peptide FDR per experiment : %
but overall peptide FDR across all 36 experiments: ~7.2%.
Changes will seek to bring the overall peptide FDR’s down to ~1%
require multiple observations (protein, P-site) across experiments
raise score thresholds
Quantitation
Will use PIP(precursor ion purity) filtering to exclude from quantitation but not identification.
PIP > 50% excludes ~7.8% of spectra.
Filtering reduces standard deviations on protein & phosphosite level iTRAQ ratios

17. Transcript present in 18/40 Common Control Members
Y Chromosome Frameshift - CD99 antigen Observed in 36 Proteomics Experiments
E159
Partial exon deletion splice, plus frameshift truncation
Max RNA-Seq Reads: 12
Transcript present in 18/40 Common Control Members

18. Acknowledgments Broad Institute/FHCRC Steve Carr Karl Clauser
Michael Gillette
Jana Qiao
Philipp Mertins
DR Mani
Eric Kuhn
Sue Abbatiello
Amanda Paulovich
Pei Wang
Sean Wang
Ping Yan
Washington U./MD Anderson/NYU
Sherri Davies
Matthew Ellis
David Fenyo
Kelly Ruggles
Reid Townsend
Li Ding
NCI Staff
Emily Boja
Mehdi Mesri
Rob Rivers
Chris Kinsinger
Henry Rodriguez
Funding
National Cancer Institute

19. Single AA Variants may be Somatic in Some Patients, Germline in Others
Nov 2013
Genomic
Highly Interesting, should correlate with prognosis and/or subtype.
May correlate with prognosis?
Might as well be canonical isoforms?
Detectable, but too rare to indicate biology.
Proteomic
G&S mix genomic variants have the highest observation rate by Proteomics.
Genomic variants present in only a single patient are observable by Proteomics

20. Not all Germline &Somatic mix Single AA Variants are “Essentially” Germline
81 Patients
Nov 2013
Genomic
Proteomic
Is G&S mix status primarily an artifact of variant calling accuracy/sensitivity?
Is there some cancer biology involved for high S/G ratio variants?
Are patients with germline form more cancer prone?
Does somatic form correlate with prognosis, development of drug-resistance?

21. Wide Range of Somatic Single AA Variants/Patient
Skip
Low confidence thresholds applied to calls
Variants: >2 QUAL score (phred-scaled)
Alternative splices: >1 read