Biology Reference
In-Depth Information
Unfiltered 100% matched features (842)
a
b
5.5
pH 4
pH 7
~200 kDa
virulent
cytoplasmic
membrane
non-virulent
~10 kDa
−
5.5
−
20
20
PC1: 80.3% of variance
Fig. 1. Principal component analysis (PCA) showing a high signal-to-noise ratio. The cytoplasmic and membrane fractions
from two related strains of
Helicobacter pylori
that differ in carcinogenic potential were analyzed by DIGE. Protein lysates
were extracted and fractionated from the B128 (nonvirulent) and 7.13 (virulent) strains independently in quadruplicate, and
842 features were matched across all eight gels. (
a
) Representative DIGE gel from an 8-gel set used to coresolve the result-
ing 16 individual samples (labeled with either Cy3 or Cy5 using a dye-swapping strategy) along with the Cy2-labeled
mixed-sample internal standard. (
b
) PCA was performed on the unfi ltered dataset. The protein expression characteristics
from 842 features from each individual sample are represented by each of the 16 data points in the PCA score plot. This
analysis demonstrated that 80.3% of the variance (PC1) separated cytoplasmic from membrane samples as expected.
An additional 5.2% of variance (PC2) separated the virulent strain 7.13 from the nonvirulent strain B128. Adapted from ref. (
6
) .
3.5. Example 2: Low
Signal, Low Noise
In another
H. pylori
study, Loh et al. investigated differential
protein expression between wild-type and mutant strains deleted
for the ArsS component of the ArsRS signal transduction system in
response to growth in different pH media (
7
). The experimental
design was similar to that described for Example 1, with
N
= 4
independent (biological) replicates from two strains grown at two
pH conditions, resulting in 16 samples coresolved across 8 DIGE
gels, each of which contained an aliquot of a Cy2-labeled mixed-
sample internal standard.
Six hundred and thirty-nine features were matched across all
eight DIGE gels, all of which was evaluated by PCA. Using no
missing values in the data (100% matching), PC1 accounted for
56.7% of the variation among these features and organized the
samples by genotype, with none of the other principal components
organizing the samples based on pH treatment (see Fig.
2
). Thus,
the biological signal with respect to genotype is high, but any signal
consistent with pH-specifi c growth was too low to be visualized
over the genotypic signal or any other technical noise present in
the overall variation.
Search WWH ::
Custom Search