effectively dealing with transition selection and data analysis for multiplexed quantitative srm-ms...
TRANSCRIPT
Effectively Dealing with Transition Selection and Data Analysis for Multiplexed
Quantitative SRM-MS Assays across Multiple Vendor Instruments
Susan Abbatiello, Ph.D.Skyline User’s Meeting
May 20, 2012
CPTAC – Clinical Proteomic Technologies Assessment for Cancer
NCI established CPTAC October 2006 to Support Biomarker Development
• Evaluate and standardize proteomic verification platforms for analysis of cancer-relevant proteomic changes in human clinical specimens.
Endogenous 12Csignature peptides
Spike heavy (13C6)-labeled peptides
Define “Signature peptides” for candidate biomarkers
Synthesize 13C/15N-labeled versions of signature peptides for use as internal standards
Candidate Protein Biomarkers
Is SID-MRM-MS Technology Reproducible, Transferrable, and Sensitive? Yes!
MRM-MS
• Observed ratio gives precise, relative quantitation across samples
• 10’s to 100’s peptides can be simultaneously quantified
• Observed ratio gives precise, relative quantitation across samples
• 10’s to 100’s peptides can be simultaneously quantified
Ratio 13C-peptide to 12C-peptide by SID-MRM-MS
12C-peptide analyte
13C6-peptide “heavy standard”
Time, minutes
Ab
un
da
nce
Whiteaker, et al, JPR 2007……………….Keshishian et al, MCP, 2007 and 2009….Hoofnagle et al, Clin. Chem. 2008……….Addona et al, Nat. Biotech. 2009…...……Kuhn et al, Clin Chem 2009………………
Williams et al, JPR 2009…………………
Ossola et al, Methods Mol. Bio., 2011…..Selevsek et al, Proteomics, 2011………..
Breast cancerCardiovascular markersThyroglobulinInterlab studyIL-33, Troponin IC-Reactive ProteinGlycated peptidesUrine proteins
– Especially with Skyline!
Establish Instrument Specific Ranges for
o RT Variability
o Peak Area
o Peak Width
o Carry over
o Column conditioning
Study 9S: Participants, Platforms, and Objectives
Prior to analyzing complex samples, are LC-MRM-MS systems running in optimal condition?
Michrom Mix6 bovine proteins, digested
50 fmol/uL
Site 52, ABI 4000 QTRAP
Site 56, ABI 4000 QTRAP
Site 56A, ABI 5500
Site 56B, Agilent 6460 ChipCube
Site 73, ABI 4000 QTRAP
Site 32, ABI 4000 QTRAP
Site 90, Agilent 6410 ChipCube
Site 98, ABI 4000 QTRAP
Site 86, ABI 4000 QTRAP
Site 86A, Waters Xevo
Site 65, Thermo Vantage
Site 19, ABI 4000 QTRAP
Site 19A, Agilent 6410 ChipCube
Site 20, Thermo TSQ Quantum
Site 95, ABI 4000 QTRAP
Site 54, ABI 4000 QTRAP
GO No GO
Define Pass/Fail Criteria
12 Laboratories4 MS Vendors7 MS models5 LC models
Development of a System Suitability Protocol for Multiple Instrument Platforms
Output: Spectral Library
Input: DDA Search results
(mzXML, pepXML, etc)
Output: Vendor Specific MRM
Instrument Method
Input: Targeted Peptide List
Output: Peak identification and
automatic integration
Input: Vendor specific acquisition files
Output: Comprehensive
results report
Input: Integrated peaks, report parameters
Selection of 22 target peptides
MRM dataacquisition
Userdata
analysis
ExternalCalculations:
RT ViewerR Scripts
Tools were created to handle workflow and data
List of 9 final peptides for evaluation
Problems Can Be Visualized Early:Peak Area Stability in Skyline
Peak area stability over 10 replicates Site Z
TA
A (1
4.5
)
YS
T (
14.9
)
GF
C (
16.0
)
HLV
(1
7.0
)
YN
G (
18
.1)
DG
G (
18.2
)
VLV
(19.7
)
VLD
(21
.0)
HG
G (
22.1
)
CA
V (
23.0
)
LVN
(2
3.2
)
YN
L (
23.5
)
IVG
(2
5.0
)
SLH
(25
.3)
DD
G (
26.8
)
IHG
++
(28.0
)
IHG
++
+ (
28.0
)
CV
A (
28.1
)
LG
E (
28.9
)
LS
F (
31.3
)
YLG
(37.9
)
VG
P (
38
.2)
FF
V (3
9.0
)
Peptide
0
10
20
30
40
Peak
Are
a (
10^
6)
Low peak areafor late eluting peptides
before
TA
A (
14
.5)
YS
T (
14.9
)
GF
C (
16
.0)
HLV
(17
.0)
YN
G (
18.1
)
DG
G (
18.2
)
VLV
(19.7
)
VLD
(21.0
)
HG
G (
22.1
)
CA
V (
23.0
)
LVN
(23
.2)
YN
L (
23.5
)
IVG
(25
.0)
SLH
(25.3
)
DD
G (
26.8
)
IHG
++ (
28
.0)
IHG
+++
(28.0
)
CV
A (
28.1
)
LG
E (
28.9
)
LS
F (
31.3
)
YLG
(37.9
)
VG
P (
38.2
)
FF
V (
39
.0)
Peptide
0.0
0.1
0.2
0.3
0.4
0.5
Peak A
rea C
V
Elevated area cv’s for late eluting peptides
0.3
before
Peak area CV over 10 replicates Site Z
Pea
k A
rea
CV
Pea
k A
rea
(10E
6)
after
TA
A (
20.4
)
YS
T (
20.8
)
GF
C (
22.0
)
HLV
(2
2.9
)
YN
G (
24
.0)
DG
G (
24.2
)
VLV
(2
5.8
)
VLD
(27
.0)
HG
G (
28.1
)
CAV
(2
8.8
)
LV
N (
29.3
)
YN
L (
29.6
)
IVG
(30
.7)
SLH
(30
.8)
DD
G (
32.3
)
CVA
(33.4
)
IHG
++
(33.5
)
IHG
++
+ (
33.5
)
LG
E (
34.6
)
LS
F (
36.8
)
YLG
(42.9
)
VG
P (
43
.2)
FF
V (
44.7
)
Peptide
0.0
0.1
0.2
0.3
0.4
0.5
Pe
ak A
rea
CV
after
CVs < 0.10
TA
A (
20.4
)
YS
T (
20.8
)
GF
C (
22.0
)
HLV
(22
.9)
YN
G (
24.0
)
DG
G (
24.2
)
VLV
(25.8
)
VLD
(27
.0)
HG
G (
28.1
)
CA
V (
28.8
)
LVN
(29
.3)
YN
L (
29.6
)
IVG
(30.7
)
SLH
(30
.8)
DD
G (
32.3
)
CV
A (33
.4)
IHG
++ (
33.5
)
IHG
+++
(33
.5)
LG
E (
34.6
)
LS
F (
36.8
)
YLG
(42.9
)
VG
P (
43.2
)
FF
V (
44.7
)
Peptide
0
10
20
30
40
Peak A
rea (
10^
6)
afterImproved peak area
for late eluting peptides
Pea
k A
rea
CV
Pea
k A
rea
(10E
6)
unlabeled
protein
15N labeled protein
Trypsin
34 proteins
depleted plasma
+
13C/15N labeled peptide
s
Fixed Spike Level
Varying Concentration
s
27 proteins
LC-MRM-MS
125 peptides
Fixed Spike Level
y6 y8y7
Recovery from Assayy8y6 y7
Figures of Merit (LOD, LOQ)
y8y6 y7
Goals:• Demonstrate cancer relevancy• Prove feasibility of > 100-plex (34 proteins) assays in plasma• Improve LOD and LOQ by depleting abundant proteins• Demonstrate true quantitative accuracy and evaluate depletion/digestion
recovery using heavy labeled proteins• Conduct blinded verification study to assess accuracy, precision and
reproducibility across multiple sites and instrument platforms• Evaluate system suitability test in context of this large-scale inter-lab study
34 proteins, 1095 transitions, 9 participating sites, 14 instruments, 4 vendors
CPTAC VWG Study 9 – Targeting 34 Proteins in Depleted Plasma, 125 Peptide Targets
Peptide and Transition Selection is Streamlined using Skyline
Lys-C/Trypsin
+
Selection of
123 target
peptides
Top 5 Product Ions
MRM-MS Data Acquisition
• Selection of best 3 ions
• CE Calculation
Final Transition ListL/H: 750 Transitions
L/H/15N: 1095 Transitions
CE publication: B. MacLean et al, 2010, Anal Chem
DDA LC-MS/MS, Database
searchSpectral Library
Spectral Libraries Focus Peptide and Transition Selection
Retention Time Scheduling:A Necessity for >100 Transitions
Inj. 1 Inj. 2 Inj. 3
28.3 min 28.2 min 28.0 min Scheduling puts rigorous demands on RT reproducibility
Peak width and RT drift are often limiting factors
Different peptides shift to various degrees.
# C
on
curr
ent
Tran
siti
on
s
80 trs
190 trs
325 trs Large numbers of transitions
require narrow RT windows or longer cycle times
Cycle times may be governed by chromatographic peak width
Skyline helps gauge number of concurrent transitions based on RT window
2 min
5 min
10 min
Retention Time Scheduling for 1095 Transitions is Challenging – and Different from System to System
3x10
0
0.25
0.5
0.75
1
1.25
1.5
1.75
2
2.25
2.5
2.75
3
3.25
3.5
3.75
4
4.25
4.5
4.75
5
5.25
5.5
5.75
6
6.25
6.5
6.75
7
7.25
7.5
7.75
8
8.25
8.5
8.75
Cpd 1: HLTASEAK: +ESI MRM Frag=380.0V [email protected] (434.3 -> 727.4) 043012_Study9-2_Site56B_A1_Calcurve_run_025.d
1 1
Counts vs. Acquisition Time (min)1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
Agilent 6490 ChipCube
AB Sciex 4000 QTrap
Data Quality Filtering and Custom Annotation by Operators for Data Sets Improves LOD
Automated version = “AuDIT”
Flags potentially bad transitions• poor peak shape• interferences• missing data
Reduces manual inspection to questionable data
Reduces subjectivity in data analysis(Abbatiello, Mani et al. Clin. Chem. 2010)
custom annotation
raw data
44 amol/uL 23 amol/uL
Automatic Integration as Good as Manual Intervention (but takes less time)
p = 0.6
Pre-process & Filter
Concentration & sample info
QuaSAR Overview:Quantitative Statistical Analysis of Reaction Monitoring Results
Overall Reproducibility
Poster ThP12, #284
10
11
12
13
14
1
2
3
4
5
6
78
9
Outcome of CPTAC Study 9 is Promising for the Use of Highly Multiplexed SID-MRM-MS Assays
Median CV at each Concentration, Study 9.1
LOD Distribution for all peptides across Sites, Study 9.1
1 2 3 4 5 6 7 8 9 121110 13 14
Site Number
0 0.001 0.075 0.32 5.61.3 24 1000.010 0.018
Concentration of each peptide (fmol/mL in 0.5 mg/mL depleted plasma)
10
11
12
13
14
1
2
3
4
5
6
7
8
9
Good Reproducibility and Accuracy is Demonstrated Through Blinded Samples
1 2 3 4 5 6 7 8 9 121110 13 14
72 fmol/mL19 fmol/mL
1.8 fmol/mL
0.1 fmol/mL
Blinded Sample Concentration Distribution
Site
Me
an
De
term
ine
d C
on
cen
tra
tion
(fm
ol/m
L)
Blinded Sample %CV Distribution
1 2 3 4 5 6 7 8 9 121110 13 14Site
Inter-Lab CV45%17%15%16%
15N Protein Standards Improve Quantitative Accuracy
Peptide Conc(15N)
Light Peak Area
15N Peak Area
Peptide Conc(13C/15N)
Light Peak Area
13C/15N Peak Area
13C/15N Peptide Internal Standards
15N Protein Internal Standards
A B
mL
x10 fmol
mL
x= =
Transition % Recoveryy5 55y6 57y8 50
Transition % Recoveryy5 110y6 111y8 94
1.3 fmol/mL concentration point
1.3 fmol/mL concentration point
25 fmol
Protein Digestion Controls Help Gauge Assay Variability
Light Peak Area from Protein Digestion Controls
13C/15N Peak Area from post-desalt peptide spikes
Pe
ak
Are
a C
V(%
)Process Variability:
Technical Variability:
Technical and Process Variability Assessed From Digestion Controls and SIS Peptide Spikes
Aprotinin 1
C-reactive protein 2
Horseradish peroxidase 1
Leptin 1
Myelin basic protein 2
Myoglobin 1
Control Proteins# Peptide Targets
Poster MP01, #004
Skyline Facilitates Rapid Data Analysis Through Overview Plots
Peak Area Replicate View, Light and Heavy
Peak Area CV Plots Provide Quick Assessment of Reproducibility Across a Series of Samples
Retention Time Reproducibility Plots Show Trends in Retention Time
Quick View of All Replicates
Interference Visualization
Heavy Peptide Transitions
Light Peptide Transitions
Summary
• First large-scale interlab study to include 15N protein reagents and >100 peptide targets (>350 peptide forms)
• Sensitivity improvement from previous study by using depleted plasma, adjusting the gradient
• Transition selection and MS method transfer across 4 instrument platforms facilitated through Skyline
• Peak Area and Retention Time views help quickly assess data quality
• Customizable reports from Skyline enable down-stream processing, helps remove subjectivity of data evaluation, and increases data analysis throughput
• Skyline helps maintain objective processing of data, requiring less manual tweaking
• It’s free, it is easy, and it will process your data
CPTAC VWG Participants & Acknowledgements
NISS: Xingdong Feng, Nell Sedransk, Jessie XiaNIST: Paul RudnickNew York University: Thomas A. Neubert, Åsa Wahlander, Sofia Waldemarson, Pawel Sadowski, John LyssandPlasma Proteome Institute: N. Leigh Anderson Purdue University: Charles Buck, Fred Regnier, Dorota Inerowicz, Vicki HedrickUniversity of California, San Francisco:Simon Allen, Susan J. Fisher, Steven C. Hall,University of North Carolina: David RansohofUniversity of Victoria: Christoph H. Borchers, Angela Jackson, Derek SmithUniversity of Washington: Michael MacCoss, Brendan MacLean, Daniela TomazelaVanderbilt University: Daniel Liebler, Kent Shaddox, Corbin Whitwell, Lisa Zimmerman
Broad Institute: Susan Abbatiello,Terri Addona, Steven A. Carr, Hasmik Keshishian, D.R. Mani, Michael Burgess, James MarkellBuck Institute for Age Research:Michael P. Cusack, Bradford W. Gibson.Jason M. Held, Birgit SchillingFred Hutchinson Cancer Research Center: Amanda G. Paulovich, Jeffrey R. Whiteaker, Shucha ZhangIndiana University: Mu Wang, Jong-Won Kim, Jimsan You Massachusetts General Hospital:Steven J. SkatesMemorial Sloan-Kettering Cancer Center: Paul Tempst, Mousumi GhoshNational Cancer Institute: Emily BojaTara Hiltke, Christopher Kinsinger,Mehdi Mesri, Henry Rodriguez, Robert Rivers
Funding: National Cancer Institute
Skyline…So easy a baby can do it
1000 Q1/Q3 Pairs – AB Sciex 4000 QTRAP
334 precursors:
108 peptides in 3 forms
10 control peptides
Gradient Optimization will Improve Sensitivity and Data Acquisition
RT: 0.00 - 80.00
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80Time (min)
0
20
40
60
80
100
0
20
40
60
80
100
0
20
40
60
80
100
Re
lativ
e A
bu
nda
nce
0
20
40
60
80
100
15.18474.1861
17.39632.5770
19.41613.9241
20.92574.5377
55.49354.2199
52.76352.995111.73
464.3190
15.95515.9611
16.52365.4940
15.59474.5050
17.18539.0721
22.44497.0310
32.98583.0168
14.51464.3874
28.38614.0621 55.99
707.808634.70
489.640452.77
344.2659
53.96354.0204
55.28354.2427
16.68474.3280
56.64353.9474
17.22527.0361
28.84632.3549
53.15353.782933.55
696.008836.36
613.907940.88
583.004724.54366.8184 52.18
354.038514.01
464.380743.77
558.9189
17.00526.8428 28.67
632.428917.83
529.8265 33.12696.0186
24.20596.3886 55.41
354.098438.03
546.870853.76
354.060440.05
583.048011.40
464.325956.71
354.140352.86354.083115.20
418.103442.93
559.1557
NL: 1.86E6Base Peak F: ITMS + c NSI E Full ms [300.00-1500.00] MS E051110_Pool_Study7grad_03
NL: 1.64E6Base Peak F: ITMS + c NSI E Full ms [300.00-1500.00] MS E0506010_Pool_Grad1_R2_03
NL: 4.65E5Base Peak F: ITMS + c NSI E Full ms [300.00-1500.00] MS E0507010_Pool_Grad2_R2_03
NL: 1.09E6Base Peak F: ITMS + c NSI E Full ms [300.00-1500.00] MS E051110_Pool_Grad3_07
(fmol/µL)2501135123104.62.00.9
0.420.190.090.04
0.0170.0080.0040.002
LOD/LOQ Calculations: How Many Points in the Curve are Needed?
What is the ideal concentration range?
LOD = sblank + t0.95 x (sblank + slow)/√n
Keshishian et al, (2009) MCP
0.001 0.01 0.1 1 10 1000.001
0.01
0.1
1
10
100
Linnet & Kondratovich, (2004) Clin Chem
LOQ range
LODrange
Proposed:
• Generate preliminary curves (16 pts)
• Pick a range and number of points to cover most peptides
16 Point Curve at Selected CPTAC Sites Shows Good Reproducibility and Sensitivity
Median LOD(fmol/uL) 0.22 0.032 0.17 0.13 0.055 0.027 0.044 0.023
Outliers 15 16 15 15 15 12 23 19
LOD is Highly Dependent Upon System Performance: Chromatography and Ionization
4000 QTRAPMedian LOD: 32 amol
QTRAP 5500Median LOD: 220 amol
Pre-Assay System Suitability Runs (5)
25% CVs25% CVs
Throughout Assay System Suitability Runs (24)
30% CVs70% CVs
Unstable ESI was a major factor in poor detection and reproducibility
System Suitability assessment detects poor system performance