Biomedical Engineering Reference
In-Depth Information
Fig. 2.
Part of the mass-frequency spectrum. The mass range is from 4950Da to 5994Da.
to remove the \noisy" bins. Again, the cut-o level can be an empirical
value or it can be determined by statistical estimation. In our procedure,
we use the cut-o level of 5%. That is, if the number of samples is 100, then
the bins with 5 peaks or less will be considered as invalid bins.
Experimental results show that PSB reaches about the same goal as
GAB does. In GAB, the initial step and crossover operations contain ran-
dom factors. Therefore, the result can not be exactly reproduced. Compared
with GAB, PSB has at least the following two advantages: First, PSB con-
sistently generates the same bins on a given dataset, while GAB creates
slightly dierent bins in each run. Second, PSB is more ecient than GAB.
In particular, when the size of spectra increases, time for PSB consumes
has little change, while time for GAB increases vastly. Figure 3 shows the
bins, generated by PSB, with the mass peak distribution of the spectra.
We plot the peaks as dots in the graph and each row as a representive of
a spectrum. The x-axis is mass value and y-axis is the labeled spectrum
number.
The proposed PSB method gives a fast and accurate binning process
for high-throughput mass spectrometry data. It organizes and expresses
MS peak data in an innovative way, which makes binning process simpler
and easier. It performs well with proles from dierent research projects
Search WWH ::




Custom Search