Study Design in DIGE-Based Biomarker Discovery - Difference Gel Electrophoresis (DIGE): Methods and Protocols - page 204

3. For example, in a DIGE-based comparison of plasma samples

from cancer patients with samples from healthy controls, it has
to be expected that only a few protein spots differ between the
experimental groups. The tumor itself and the surrounding
tissue might secrete a number of different proteins which are
normally not present in the blood fl ow of healthy subjects.
However, due to the huge dilution of these proteins by the
total plasma volume, only a few of them might be detectable in
the peripheral blood. Therefore, the p 1 is often set to 0.01 for
such experiments (indicating that about 1% of all investigated
spots might show disease-related expression changes). In con-
trast, in a proteomic comparison of tumor tissue with sur-
rounding normal tissue, it has to be expected that the portion
of spots showing disease-related expression changes is much
higher ( p 1 = 0.05 or even 0.10).

4. The true effect size can be calculated from an experimental

data set as follows:
(a) q=D /
q ... effect size, D ... difference in group means of the
expression levels, s ... common standard deviation.
In a comparison of the expression pattern of two experi-
mental groups (control and group-1), D is calculated
as follows:
(b) D = meanSA 1 − meanSA c
meanSA 1 or c is mean standardized abundance of group-1
and control, respectively.
The SA values of each spot are calculated according to the
DIGE algorithm dividing the spot volume of the sample by
the spot volume of the internal standard on the same gel
( 13 ). The DeCyder software shows the SA values of the
spot of interest on the different gels in the graph view of
the BVA module. The values can be exported using the
XML toolbox.

5. The relationship between D and q is illustrated in Fig. 2 with

the help of two examples.

6. For analysis in the BVA module, open the workspace and select

the Protein Table (PT). Then open the Protein Statistics dialog
(in the Process menu). Select the type of statistical test (inde-
pendent or dependent) according to your biological question.
Activate the Average Ratio and the Student's t test option and
defi ne the two experimental groups which should be compared.
Finally, select the FDR option (this is essential to avoid the
problem of multiple testing).
For analysis in the EDA module, open the workspace and
select the Differential Expression Analysis dialog (in the
Calculations window). Select the type of statistical test