| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Editorials |
1 Department of Biostatistics and Applied Mathematics, University of Texas, M.D. Anderson Cancer Center, Houston, TX 77030
Mass spectrometry is increasingly being applied to sift through complex protein mixtures to find biomarker patterns that can be used for diagnosis, prognosis, or monitoring of disease. Although some studies have directly examined tumor tissue, most have targeted easily accessible fluids, including urine, nipple aspirate fluid, cerebrospinal fluid, plasma, and most commonly, serum. A few studies have been performed on high-resolution instruments using a matrix-assisted laser desorption/ionization (MALDI) ion source and a time-of-flight (TOF) ion detector. The majority, however, have used lower resolution instruments and surface-enhanced laser desorption/ionization (SELDI), a variant of MALDI that relies on commercially prepared surfaces (ProteinChips®), to perform protein isolation or separation.
To date, the technology has primarily been applied to search for cancer biomarkers, with some success. It is now clear that SELDI profiling of the proteome can successfully detect anonymous proteins that are produced at higher (or lower) concentrations in the tissues or fluids of cancer patients than in healthy individuals (1)(2)(3). There is also preliminary evidence that we may be able to discover patterns that can reliably distinguish cancer patients from healthy individuals (4)(5)(6)(7).
These findings should be greeted with cautious optimism. When it has been possible to identify the protein peaks, they have often turned out to be well-known acute-phase proteins. Some authors have claimed that mass spectrometry is intrinsically limited in its depth of coverage, with a dynamic range that prevents it from being able to find low-abundance proteins (8)(9). As for classification patterns, independent studies to date have failed to find the same patterns, and no pattern has yet been validated in an independent laboratory. In one study, both sensitivity and specificity decreased significantly when samples were processed in the same laboratory after a delay of several months (10).
Reproducibility can be enhanced in several ways, by improving instrumentation, study design, experimental protocols, or analysis tools. Although improved instruments might help, the studies conducted to date with higher resolution MALDI instruments have not convincingly demonstrated that they produce more robust patterns than the SELDI approach. The need for better study design and experimental design for proteomic profiling studies has been amply demonstrated and discussed elsewhere (11)(12)(13). The Early Detection Research Network has designed, and is now conducting, a careful multiinstitutional study to develop and validate protocols to produce reproducible mass spectra (14)(15).
This brings us, finally, to analysis tools. There is, as yet, no consensus on the best methods to analyze mass spectra from serum proteomic profiling experiments. At one extreme, the National Cancer Institute/Food and Drug Administration proteomic research group performs minimal preprocessing and allows the intensity at every measured mass-to-charge (m/z) value as a potential feature for distinguishing cancer patients from healthy controls (16)(17)(18). Unfortunately, their three seminal studies suffered from design flaws that confounded technologic factors with biological contrasts (11)(12)(19); therefore, their approach cannot yet be viewed as soundly established. At the other extreme, most published studies perform preprocessing and peak detection with software from Ciphergen, the manufacturer of the SELDI instrument (20). In our opinion, the Ciphergen software is extremely conservative about calling something a peak, and its baseline correction algorithm introduces substantial bias into the estimates of the size of a peak. These algorithmic weaknesses can reduce the effective sensitivity of the instrument below its true capabilities and can hamper its reproducibility.
Several attempts have been made to improve the processing of mass spectra. Yasui and coworkers (21)(22) used a supersmoother to find peaks, but they settled on binary indicators of presence or absence in lieu of quantification. We introduced an iterative method to perform background correction and peak finding simultaneously as part of a quality-control procedure (23). Qu et al. (24) applied wavelets for data reduction, but the wavelet coefficients do not directly correspond to physical quantities and are therefore difficult to interpret. We have also used wavelets for noise reduction (25), particularly in concert with the mean spectrum to borrow strength across spectra (26). Our studies benefited from a computer model based on the physics of a mass spectrometer that used an extremely simple model of the detector (27). Sauve and Speed (28) used dynamic programming to improve calibration across multiple spectra and morphologic filters for baseline correction.
In this issue of Clinical Chemistry, Malyarenko et al. (29) introduce some exciting ideas from the theory of time series into the analysis of SELDI spectra. They first improve our understanding of baseline by realizing that its primary source resides in the physics of the ion detector, which can be saturated and decays slowly. They then use a deconvolution filter, tailored to a size corresponding to sodium adducts that commonly occur in mass spectra, to smooth the signal and obtain better resolution of individual mass peaks. They provide substantial evidence, based on spectra from both single laser shots and averages over many laser shots, that their analytical methods can improve the effective resolution of the instrument.
This report shows that incorporating knowledge of the physical and chemical properties of mass spectrometry instruments and methods can lead to improved analytical tools. The next step will be to show that better analytical tools can contribute to better reproducibility of the results of serum proteome-profiling studies.
References
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  B. H.L. Goulart, J. W. Clark, H. H. Pien, T. G. Roberts, S. N. Finkelstein, and B. A. Chabner Trends in the Use and Role of Biomarkers in Phase I Oncology Trials Clin. Cancer Res., November 15, 2007; 13(22): 6719 - 6726. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Zhang and D. W. Chan Cancer Biomarker Discovery in Plasma Using a Tissue-targeted Proteomic Approach Cancer Epidemiol. Biomarkers Prev., October 1, 2007; 16(10): 1915 - 1917. [Full Text] [PDF]
|
|
|
|
 |
|
 N. Filigheddu, V. F. Gnocchi, M. Coscia, M. Cappelli, P. E. Porporato, R. Taulli, S. Traini, G. Baldanzi, F. Chianale, S. Cutrupi, et al. Ghrelin and Des-Acyl Ghrelin Promote Differentiation and Fusion of C2C12 Skeletal Muscle Cells Mol. Biol. Cell, March 1, 2007; 18(3): 986 - 994. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. W. Randolph, B. L. Mitchell, D. F. McLerran, P. D. Lampe, and Z. Feng Quantifying Peptide Signal in MALDI-TOF Mass Spectrometry Data Mol. Cell. Proteomics, December 1, 2005; 4(12): 1990 - 1999. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. E. Banks, A. J. Stanley, D. A. Cairns, J. H. Barrett, P. Clarke, D. Thompson, and P. J. Selby Influences of Blood Sample Processing on Low-Molecular-Weight Proteome Identified by Surface-Enhanced Laser Desorption/Ionization Mass Spectrometry Clin. Chem., September 1, 2005; 51(9): 1637 - 1649. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. T. Fung and E. Gavin Analysis of Mass Spectrometry Profiles of the Serum Proteome Clin. Chem., July 1, 2005; 51(7): 1309 - 1309. [Full Text] [PDF]
|
|
|
|
|
|
K. R. Coombes Analysis of Mass Spectrometry Profiles of the Serum Proteome - Reply Clin. Chem., July 1, 2005; 51(7): 1309 - 1309. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
