Global Gene Expression Assays: Quantitative Noise Analysis - Systems Biology: Networks, Models, and Applications - page 105

cell line, so the measured expression value differences as shown in
figure 4.2b contain the combined effect of the genuine gene expression
differences between the two cultures together with differences caused
by measurement error. Therefore, to correctly assess the statistical
relevance of the measured gene expression differences between two
experiments, such as 1 and 10, it is crucial to characterize the fluctua-
tions caused purely by experimental measurement, such as the noise
shown in figure 4.2a.
Although experimental noise is known to be a feature of microarray
experiments, only recently has it been studied systematically by repli-
cate experiments [6,11]. In particular, for oligonucleotide microarrays,
Novak et al. [11] characterized the dispersion between two experi-
ments by the standard deviation of their corresponding gene
expression levels. Using this measure of dispersion, they studied the
different effects of experimental, physiological, and sampling variability,
which provide important guidance for microarray experiment design.
Following this methodology, this section focuses on understanding
how different experimental steps contribute to the total noise and what
the possible mechanism for the noise could be. The distribution of the
noise is studied in detail; this is used in devising a statistical method to
determine differentially expressed genes.
In order to separate the different noise sources, all the replicate
experiment pairs are placed into two groups. Group G 1

consists of all

the pairs that differ only in the hybridization step:

G 1 ={(1,2),(3,4),(5,6),(8,9),(9,10),(8,10)}

Group G 2 consists of all the replicate experiment pairs that are
carried out separately right after the extraction of the mRNA:

G 2 ={(1,3),(1,4),(2,3),(2,4),(5,7),(5,8),(5,9),(5,10),(6,7),(6,8),(6,9),
(6,10),(7,8),(7,9),(7,10)}

While gene expression differences between pairs of experiments
in G 2 represent the full experimental noise, G 1 has been constructed
to extract the noise due to hybridization alone. For reference, all
the nonreplicate experiment pairs are placed into a third group
.

Gi j {( , ), 1 ≤≤ i 4 5 , ≤ ≤ j 10 }

=

3

NOISE DISTRIBUTION
As is evident from figure 4.2, the noise depends strongly on the expres-
sion level. Therefore, an expression-dependent distribution function is
needed to characterize the variability between replicates. Given two
measured gene expression values, q 1 and q 2 , for the same gene from two