[40]. It does not need a reference sample and instead calculates
the relative intensities of every protein for each particular label
in relation to the sum of intensities obtained for that protein in
all labels (Fig. 2B). Note that in either of these interexperiment normalization methods, it is necessary to run first
intra-experiment normalization as described above.
3.10 Statistical
Analysis: Differentially
Expressed Proteins
Once proteomic data set with normalized values is ready, different
parametric and nonparametric test analyses can be carried out.
Before running parametric tests, the usual procedure for proteomic
data is to logarithmically transform the normalized protein intensities/signals in order to meet normality and variance homogeneity
assumptions required for parametric testing. At this point, it is
important to consider the multiple hypothesis testing (multitest)
problem that has been extensively explained and discussed in Diz
et al. [41]. A new strategy to deal with this problem has been
suggested in this cited work, consisting of applying one or several
correction methods (for instance, combining the information
obtained from SFisher and q-value). In order to follow this suggestion, different multitest corrections can be calculated and applied
easily to an initial list of p-values using the spreadsheet provided as
supplementary material (unblock file after downloading by rightclick file/properties) in Diz et al. [41]. Calculation of q-values [42]
can be carried out in http://qvalue.princeton.edu or using the
qvalue function available in R package (https://github.com/
StoreyLab/qvalue). Thus, a new column with calculated q-values
can be added to the same spreadsheet that includes results related to
other multitest correction methods for comparative purposes. The
idea behind this strategy is that final decision about whether using a
more or less stringent correction method could depend on the
nature of each research study. For some studies, mainly of
exploratory-type, the level of stringency for controlling the false
positive rate could be lower than that applied in other studies that
are mainly of confirmatory type. In the former scenario, applying a
less stringent correction method such as SFisher could be sensible,
while in the latter scenario, the use of a more stringent method such
as sequential Bonferroni could be necessary especially in cases
where one cannot afford to provide any false discovery.
4 Notes
1. Instructions about how to correctly set up the SDS-PAGE
equipment (Mini-PROTEAN
® Tetra Vertical Electrophoresis
Cell, Bio-Rad Laboratories, USA), polyacrylamide gels, and
different buffers (sample and running buffers) are available in
the pdf of Mini-PROTEAN
®
Tetra Cell Instruction Manual
that can be download from http://www.bio-rad.com.
94
Angel P. Diz and Paula Sa ´ nchez-Marı ´n
the relative intensities of every protein for each particular label
in relation to the sum of intensities obtained for that protein in
all labels (Fig. 2B). Note that in either of these interexperiment normalization methods, it is necessary to run first
intra-experiment normalization as described above.
3.10 Statistical
Analysis: Differentially
Expressed Proteins
Once proteomic data set with normalized values is ready, different
parametric and nonparametric test analyses can be carried out.
Before running parametric tests, the usual procedure for proteomic
data is to logarithmically transform the normalized protein intensities/signals in order to meet normality and variance homogeneity
assumptions required for parametric testing. At this point, it is
important to consider the multiple hypothesis testing (multitest)
problem that has been extensively explained and discussed in Diz
et al. [41]. A new strategy to deal with this problem has been
suggested in this cited work, consisting of applying one or several
correction methods (for instance, combining the information
obtained from SFisher and q-value). In order to follow this suggestion, different multitest corrections can be calculated and applied
easily to an initial list of p-values using the spreadsheet provided as
supplementary material (unblock file after downloading by rightclick file/properties) in Diz et al. [41]. Calculation of q-values [42]
can be carried out in http://qvalue.princeton.edu or using the
qvalue function available in R package (https://github.com/
StoreyLab/qvalue). Thus, a new column with calculated q-values
can be added to the same spreadsheet that includes results related to
other multitest correction methods for comparative purposes. The
idea behind this strategy is that final decision about whether using a
more or less stringent correction method could depend on the
nature of each research study. For some studies, mainly of
exploratory-type, the level of stringency for controlling the false
positive rate could be lower than that applied in other studies that
are mainly of confirmatory type. In the former scenario, applying a
less stringent correction method such as SFisher could be sensible,
while in the latter scenario, the use of a more stringent method such
as sequential Bonferroni could be necessary especially in cases
where one cannot afford to provide any false discovery.
4 Notes
1. Instructions about how to correctly set up the SDS-PAGE
equipment (Mini-PROTEAN
® Tetra Vertical Electrophoresis
Cell, Bio-Rad Laboratories, USA), polyacrylamide gels, and
different buffers (sample and running buffers) are available in
the pdf of Mini-PROTEAN
®
Tetra Cell Instruction Manual
that can be download from http://www.bio-rad.com.
94
Angel P. Diz and Paula Sa ´ nchez-Marı ´n