Biology Reference
In-Depth Information
variable importance plots (VIP).
20
Metabolites
with VIP values more than 1.0 are considered
signi
baseline correction, noise reduction, deconvolu-
tion, peak area calculation and retention time
alignment are performed to extract mass spectro-
metric and chromatographic information from
all analyzed samples into a single data table
using GC/MS vendor software packages such
as ChromaTOF or external programs such as
XCMS and MZmine. Data cleanup to remove
artifact peaks or peaks with poor repeatability
(e.g., detected in less than 50% QC samples or
high variability with coef
cant in accounting for class discrimination
between test and control.
29
Univariate statistical
tests such as Welch
s
t
-test are performed to
assess the statistical signi
'
cance of these differ-
entiating marker metabolites. The potential
identities of these marker metabolites can be
realized through matching the EI mass spectra
and retention indices (RI) with mass spectral
libraries, namely, National Institute of Standards
and Technology (NIST), Golm Metabolome
Database (GMD), or Human Metabolome Data-
base (HMDB). More details regarding RI are
presented in the next section. De
cient variation [CV]
more than 30% in QC) is performed for quality
assurance.
19,20
Normalization is done to remove
systemic variation in the data due to change in
instrument response during the course of anal-
ysis or the effect of varied metabolite dilution
in urine sample. Subsequent to preprocessing,
the data is processed using multivariate sta-
tistical methods. Multivariate analysis can be
classi
nitive identifi-
-
cation of metabolites requires con
rming their EI
spectra and RI with that of pure standards
analyzed under identical analytical condition.
4
The current marker metabolite identi
cation
process is limited by the incomplete character-
ization of the metabolome and unavailability
of standard metabolites. Subsequent to the
discovery phase, further clinical validation of
the marker metabolites is imperative to prove
their true values as biomarkers.
30,31
ed into unsupervised and supervised
methods. Unsupervised multivariate analysis
such as principle component analysis (PCA)
allows the visualization of grouping trends and
inspection of outliers in data and does not use
class information (e.g., diseased versus healthy).
Supervised multivariate analysis such as partial
least square discriminant analysis (PLS-DA)
and orthogonal PLS-DA (OPLS-DA) maximizes
separation between classes of observations
based on their class information and is typically
performed after a relatively distinct separation
between groups is observed in unsupervised
analysis. PLS-DA is used to identify differenti-
ating metabolites characterizing the respective
test and control classes. Internal and external
model validation is performed subsequently to
evaluate the validity of the constructed model.
Several excellent references on metabonomic
data analysis are available.
19,20,27,28
Strengths and Limitations of GC/MS
Although GC/MS is an excellent tool for the
separation, detection, and quanti
cation of
a large number of metabolites, it remains impor-
tant for metabonomic scientists to appreciate the
limitations of GC/MS (
Table 1
). Such aware-
ness of the limitations allows scientists to plan
the metabonomic experiments optimally.
Applications
GC/MS has been widely applied in metabo-
nomic research to identify biomarkers and
elucidate disease mechanisms related to gastro-
enterological diseases,
32
central nervous disor-
ders,
12
cancers,
33
e
37
and kidney diseases.
38
Inspired by a study that demonstrated that
well-trained dogs
Biomarker Discovery
The next step in metabonomics is the
screening of biomarkers. Putative marker metab-
olites
responsible
for
class
separation are
screened and identi
ed using loadings and
could differentiate urine
Search WWH ::
Custom Search