INCOGEN 2003 Overview and Look Ahead

1
Enhancement of Sensitivity and Resolution of
SELDI-TOF MS records
ENAR 03/05
WM, INCOGEN, EVMS Collaboration
Hardware Issues
Performance Enhancement Tools
2
SELDI Profiling Sources of Error
source and detector related DSij,
Dj (DM/z)
3
Typical VIBE-MS Analysis Pipeline
4
Enhancing Data Quality

• characterizing uncertainties

• Relevance (protein?) noise vs. signal
  characteristics
• Location (Dj?) alignment, deconvolution
• Intensity (DSij?) baseline, resampling,
  deconvolution

5
Comprehensive Data Survey
6
Hardware Settings
7
SELDI Surface Evolution

• Acquisition
• requires
• optimization
• TOF to M/z
• calibration
• may vary
• Process
• individual
• shots

8
Overload Baseline

• Length is proportional to the number of overload
  points
• Single shot correction only

9
Charge Accumulation Baseline

• Baseline is predictable
• from 3-parameter
• RC-modeling
• (Clin.Chem. 2005, 51 65-74)
• Single shot or average

10
Peak Alignment Between Spots
QC record

• Detect shifts in time
• Align in time

Time point
11
TOF-Jitter Within Spot

• TOF and intensity varies
• due to surface roughness

DTOF 2 ns/mm (2 mm/dwell period) /-10 mm ?
/-5 time points
12
Subposition De-Jittering

• Correct by subposition de-jittering during
  acquisition
• Gain resolution (even from only 12 subpositions)

13
Deconvolution With Shaping Filter
14
SELDI vs. Isotopic Resolution

• Instrumental st const (lt 12 kDa) in time
• Isotopic st 2 time points (2-70 kDa)

15
Effect of Resampling on TOF Signal

• For all peaks
• st const(t)
• Enhanced
• sensitivity
• for heavy ions

16
Single Shot Noise Characteristics

• 8-bit ADC
• (max SNR 30)
• 2-point clock
• feed in FFT
• (high power)
• 2-channel
• differential
• amplifier
• (colored noise)

17
Average Noise and Resampling

• Gaussian noise
• variance is
• increased
• 200-shot average
• variance is
• preserved

18
Broad Range Deconvolution
19
Conclusions

• Know hardware to design data processing
• - Baseline models
• - Dejittering and alignment
• - Resampling and deconvolution
• Report uncertainties for data matrix
• Comprehensive data analysis with VIBE-MS

20
Acknowledgements
This work was supported by Virginias
Commonwealth Research Technology Fund IN2003-04,
and Phase II SBIR grant from the National Cancer
Institute CA101479 We thank our collaborators
and coauthors from EVMS, WM and INCOGEN for
assistance with acquisition and analysis of
calibration data for SELDI PBSII instrument We
are grateful to Dr. Stacy Moore and Dr. Scott
Weinberger from Ciphergen, Inc. for their help in
clarifying instrumental specifications and
parameters