9 Genomic Techniques and How to Apply Them to Marine Questions
361
9.4.1.2 How Many Replicates?
It is important to note that any gene expression measurement pipeline will
exhibit biological variation, not only those using microarrays. Any analysis should
therefore contain biological replicates. The harvesting of cells to extract sufficient
amounts of RNA from the organism can be particularly complicated for microarray
experiments, especially when working with eukaryotes. The necessary number of
replicates required in an experiment is thus an important question.
Unfortunately, there is no simple answer. Three replicates, however, is an often
cited quantity (Lee et al. 2000, Yang and Speed 2002), and could serve as the rule
of thumb for the lowest possible number. There also seems to be a growing trend
for journals to require a minimum of three biological replicates for publication as
well as experimental validation of results by a different technique. However, the
suitable number of replicates depends on several factors. The biological variability
is the most important, together with the size of the effect which the observer expects
to observe. If variability is high, the number of replicates should be increased.
Similarly, the smaller the measured effect, the more replicates are required to detect
it. The required precision of the study is another important factor. If the experiment
is aimed to discover a few candidate genes for further studies, it might be justified
to reduce the number of replicates.
So called power analysis methods can serve to calculate the approximate number of replications necessary. They assist in experimental design by providing an
estimate of the number of replicates, given the desired power (the ability to detect
a large proportion of the differentially expressed genes), the confidence level, and
the variability of the data – see for example Pan et al. (2002), Black and Doerge
(2002), Li et al. (2005), and in particular Page et al. (2006) who have implemented
the PowerAtlas software for power analysis based on publicly available data.
9.4.2 Gene Expression Analysis
The pipeline used for analysing the data obtained from a microarray experiment
depends on the purpose of the experiment and the experimental question. Despite
this, common analysis steps can be identified which are based on the features of the
data. Analysis steps, which are often found in the literature, will be detailed in the
following paragraphs.
9.4.2.1 Image Analysis
Data analysis of typical microarray experiments starts with the analysis of image
data generated by the scanner software. Some novel microarray platforms provide
direct signal readout via electro-chemical reactions. For these platforms, the image
analysis step is not necessary. For approaches that do require image analysis, images
for each channel are segmented. In other words, the locations of the features on
the surface need to be identified. Most image analysis software can be calibrated
361
9.4.1.2 How Many Replicates?
It is important to note that any gene expression measurement pipeline will
exhibit biological variation, not only those using microarrays. Any analysis should
therefore contain biological replicates. The harvesting of cells to extract sufficient
amounts of RNA from the organism can be particularly complicated for microarray
experiments, especially when working with eukaryotes. The necessary number of
replicates required in an experiment is thus an important question.
Unfortunately, there is no simple answer. Three replicates, however, is an often
cited quantity (Lee et al. 2000, Yang and Speed 2002), and could serve as the rule
of thumb for the lowest possible number. There also seems to be a growing trend
for journals to require a minimum of three biological replicates for publication as
well as experimental validation of results by a different technique. However, the
suitable number of replicates depends on several factors. The biological variability
is the most important, together with the size of the effect which the observer expects
to observe. If variability is high, the number of replicates should be increased.
Similarly, the smaller the measured effect, the more replicates are required to detect
it. The required precision of the study is another important factor. If the experiment
is aimed to discover a few candidate genes for further studies, it might be justified
to reduce the number of replicates.
So called power analysis methods can serve to calculate the approximate number of replications necessary. They assist in experimental design by providing an
estimate of the number of replicates, given the desired power (the ability to detect
a large proportion of the differentially expressed genes), the confidence level, and
the variability of the data – see for example Pan et al. (2002), Black and Doerge
(2002), Li et al. (2005), and in particular Page et al. (2006) who have implemented
the PowerAtlas software for power analysis based on publicly available data.
9.4.2 Gene Expression Analysis
The pipeline used for analysing the data obtained from a microarray experiment
depends on the purpose of the experiment and the experimental question. Despite
this, common analysis steps can be identified which are based on the features of the
data. Analysis steps, which are often found in the literature, will be detailed in the
following paragraphs.
9.4.2.1 Image Analysis
Data analysis of typical microarray experiments starts with the analysis of image
data generated by the scanner software. Some novel microarray platforms provide
direct signal readout via electro-chemical reactions. For these platforms, the image
analysis step is not necessary. For approaches that do require image analysis, images
for each channel are segmented. In other words, the locations of the features on
the surface need to be identified. Most image analysis software can be calibrated