Biology Reference
In-Depth Information
multivariate approaches provides a comprehensive overview of the
data, with single protein studies and multiprotein trends, maximiz-
ing the information obtained from most of the datasets.
In this chapter, we provide a model workflow for a complete
statistical analysis of proteomic datasets (Fig.
1
), starting with some
guidelines for image analysis and peptide identification, and fol-
lowing with univariate and multivariate statistics.
2
Materials
Despite the numerous and specific image capture and processing
software tools [
6
], most of them present a limited capability for
statistical analyses. The adequate data preprocessing, the presence
of missing spots or even the fact that they do not implement error
control procedures are some of their limitations. Some authors
have used different software packages over the same images, with a
difference of up to 50 % in the quantitatively differential spots [
3
].
To avoid software-induced biases and opaque processing steps, we
recommend a dedicated effort in data processing. Here, we present
a statistical workflow for maximizing the information obtained by
quantitative proteomics. All the analysis described in this chapter
can be conducted with any existing statistical package that contains
the functions described below. We recommend, however, R, which
is a free software environment for statistical computing and graphics
[
7
] (
http://www.r-project.org/
); and/or COVAIN, a user-friendly
statistical tool [
8
].
3
Methods
The main limitation of -omics analyses is the reduced number of
replicates that are usually performed, which is always limited by the
economical and processing capacities of the laboratory. Since only
a low number of analyses can be done, biological replication is
always better than technical replication. Moreover, it is not recom-
mended to carry out a technical replication if it occurs at the
expense of biological replication [
9
]. To achieve a good statistical
power, the recommended number of biological replicates is
between 5 and 7, despite statistics can be performed from three
biological replicates [
10
].
3.1 Experimental
Design
1. Images should be obtained from a calibrated densitometer/
fluorimeter with enough exposure time to capture faint spots
without saturating most abundant spots, and with the maxi-
mum reasonable resolution. Guidelines for conducting 2-DE
and gel staining have been amply described in a number of
chapters in this topic, as well as in its previous edition.
3.2 Data
Preprocessing
3.2.1 2-DE
Electrophoresis Image
Acquisition (Spot
Abundance)
