Biomedical Engineering Reference
In-Depth Information
ble distinguishing distinctive peaks for certain spectral data. As a result, it
may combine some dierent peaks together in a wide bin range. While the
alignment (binning) algorithm of the GG procedure, or the projecting spec-
trum binning (PSB) method, can eectively identify these distinct peaks.
The basic idea of PSB is to projects all MS peaks, from the view of the
top of MS spectra, to a plane, in which one MS spectrum is one row and a
MS peak is one dot on the row. These dots represent the peak distribution
in spectra. The peak distribution has been used to determine bin location
and bin width. The results show that PSB bins peaks both eectively and
eciently (see next section for details).
If we adapt the strengths of WW and GG procedure, then we have a
sketch of the WG procedure:
Step 1 to 5: adopt the WW procedure for (1) raw data calibration, (2)
wavelet denoising, (3) baseline correction, (4) normalization, and (5) peak
detections.
Step 6: adopt the GG procedure for nal peaks alignment (binning).
In the next section, we intruduce the projecting spectrum binning
(PSB) method. PSB is an equivalent yet more ecient method for peak nal
alignment than that of the GA binning method. It gaves similar results with
less computation time.
3. Projecting Spectrum Binning Method
Now, let us discuss the cross sample alignment of MS data. For data samples
from patients, as it is mentioned above, the data rst has to be preprocessed
with the proper background subtracted, normalized, and the dierent frac-
tions combined to obtain one integrated spectrum for each patient. The
integrated spectrum is then binned or aligned so that the data for all pa-
tients in the sample is formatted in a matrix with one index representing
the patients and the other index the peaks (discrete m=z's corresponding
to the mean of the m=z of each bin).
The spectral data sets that result from MS experiments consist of the
sequentially recorded numbers of ions arriving at the detector (intensity)
corresponding to the mass-to-charge ratio (m=z) values. Although variation
occurs in MS data, the following two assumptions are commonly used in MS
data processing and analysis: (i) The peaks from a protein in the spectra
should be positioned in an extremely tight mass range; (ii) the peaks, lo-
cated in an extremely tight mass range in the spectra, should be generated
by a protein. According to these two assumptions and after the peak selec-
Search WWH ::
Custom Search