Biology Reference
In-Depth Information
TIME TRACES AND MULTIPLE COMPARISONS
:
THE MACROPHAGE DATA
Changes in expression level as a function of time are particularly
important in understanding the response of cells to a perturbation.
Suppose that the aggregate tpm of a signature is measured at
n
time
points
t
0
0 (i.e., before perturbation),
t
1
,
t
2
,...,
t
n
−1
, yielding a series of
log-tpms . If the perturbation significantly affects the
expression of that signature, then one expects a small
p
-value for at
least one of the
n
=
(
θθ
,
,...,
θ
)
t
t
t
n
0
0
−
1
1)/2 pairwise comparisons between temporal
data points. It is necessary to consider all pairwise comparisons, and
not just those between consecutive measurements, because there are
numerous instances where consecutive comparisons are not beyond
the level of significance, but those between nonadjacent time points
are [13]. A significance index (SI) for the time series of a given signature
is defined as the minimum
p
-value obtained from all possible pair-
wise comparisons within the series. An SI is considered significant if
it is smaller than some chosen threshold
p
0
. Note that the most signifi-
cant
p
-value does not necessarily correspond to the largest fold change,
as the significance of a fold change depends on the expression level.
In ref. [13], measurements were reported for the gene expression of
macrophages prior to LPS stimulation and at times
t
×
(
n
−
2, 4, 8, and 24 h
following it. For each observed signature, the SI of its time series was
computed. Often, multiple signatures were found to correspond to a
single gene (in the NCBI database). In such cases, the signature with
the lowest SI values was associated with the gene. Following this pro-
tocol, 12,657 signatures, of which 2356 (20%) underwent statistically
significant changes in expression level with SI
=
0.05, were identified.
Significant signatures corresponded to well-characterized genes [13] in
greater proportion than nonsignificant signatures, an indication that
the SI allows the identification of meaningful signals from massive
amounts of signature data.
<
COMPARISON OF NOISE BETWEEN DNA MICROARRAY
AND MPSS EXPERIMENTS
Conversion of Affymetrix Expression Values into
Transcripts per Million
Statistical variability between biologically equivalent replicate experi-
ments provides one useful means of comparing different global gene
expression assay technologies. However, before comparing the level of
noise observed from DNA microarray and MPSS replicate experiments,
it is first necessary to convert the observed signals into equivalent
units, given that the noise depends on the expression level in both tech-
nologies. MPSS bead counts are converted into transcripts per million
(tpm) values following the methods described above. The comparison
Search WWH ::
Custom Search