Biology Reference
In-Depth Information
3.4. Example 1: High
Signal, Low Noise
Helicobacter pylori
is the strongest known risk factor for gastric
adenocarcinoma, yet only a fraction of infected persons ever
develop cancer. In a recent study, Franco et al. used DIGE to quan-
tify differences between two related strains of
H. pylori
, one of
which caused only gastritis in rodents whereas the other also
induced adenocarcinoma (
6
). The virulent strain was directly
derived from the nonvirulent strain after passage through an ani-
mal, leading to the expectation that only a few key changes gave
rise to the more aggressive nature of the virulent strain. Membrane
and cytoplasmic fractions were analyzed from both genotypes, and
each was produced as four independent (biological) replicates to
control for unanticipated technical variation. The resulting 16
samples were coresolved on 8 coordinated DIGE gels along with a
Cy2-labeled internal standard using standard methods as described
elsewhere in this volume.
The technical noise of the DIGE platform has been demon-
strated to be low (
1
), and in this case the additional variation (noise)
derived from sample handling and normal biological variation
was also expected to be low because the replicates involved clonal
bacterial colonies, although the fractionation into cytoplasmic
and membrane preparations could produce unanticipated technical
variation. PCA performed on 842 features matched across all 8 gels
(no missing values) demonstrated that the majority of variation
(PC1 = 80%) was consistent with differences between cytoplasmic
and membrane fractions, as expected. However, the second princi-
pal component (PC2), which described the second greatest source
of variation between the samples, accounted for only an additional
5% variation, but this variation organized the samples derived from
the virulent vs. nonvirulent strain (see Fig.
1
). Thus, PCA clearly
indicated low technical noise among the independently derived
samples and clearly indicated that a low but signifi cant level of bio-
logical variation was correlated with virulence.
That these two largest sources of variation correlated with
expectations about the biology, rather than with technical issues
such as dye-labeling bias, sample prep number (each set was pre-
pared on separate days using different reagents), or other unan-
ticipated, nonbiological factors, enabled these investigators to
focus on the small number of virulence-related changes with high
confi dence and low expectations for false discovery, and led to a
number of signifi cant fi ndings including an amino acid substitu-
tion (cysteine-to-arginine, causing a pI shift) in a fl agellar protein
that affected motility. This result was readily detected via DIGE
coupled with mass spectrometry-based protein identifi cation
because the mutation altered the pI of the resolved, intact pro-
tein forms but otherwise did not affect the relative expression
level of the fl aA protein. As such, this alteration would most likely
have gone undetected in a peptide-base, bottom-up LC-MS/MS
strategy.
Search WWH ::
Custom Search