Biology Reference
In-Depth Information
MacOS X and Linux. Alongside its base system that provides very generic functions,
R is being continuously enriched by packages that can be developed by any
researcher. These packages are available on the CRAN (The Comprehensive R
Archive Network) or through the more specialized Bioconductor project that coor-
dinates a large number of open source R package contributions in the field of com-
putational biology (
Gentleman
et al.
, 2004
). A number of approaches, which are
often computationally intensive and therefore not well suited for interactive use with
R, are distributed as standalone applications. An alternative to R is provided by com-
mercial software such as GeneSpring (distributed by Agilent Technologies) offering
environments for microarray data analysis with more intuitive interfaces that do not
require programming skills and allow guided but yet somewhat flexible analysis
workflows.
2.2
Experimental design
When designing a microarray study, the experimental details including data analysis
procedures and, if applicable, integration with data from other ('omics) approaches
need to be dealt with carefully in order to ensure that the generated data will address
the relevant biological questions. More specifically, given the amount of work and
resources needed to perform the experiments, several issues crucial for obtaining
most informative results should be considered from the outset. These include the data
type (expression microarray vs. genomic tiling array), the experiment type (one-
colour vs. two-colour hybridizations), the number of biological replicates and the
potential batch effects (
Leek
et al.
, 2010
). Certainly, quantifying (changes in) tran-
script abundance under different experimental conditions and/or time-points is usu-
ally the main objective of a transcriptome study. An additional substantial aspect of
unbiased transcriptomics approaches is to detect and annotate transcribed regions
that lie outside the current genome annotation. Mainly because of the latter aspect,
tiling arrays often outperform classical microarrays in systems biology studies. How-
ever, in view of constraints such as costs or availability of probe design and data anal-
ysis capabilities, expression microarrays remain a useful tool for the analysis of
annotated transcripts. For a series of studies involving a single bacterial species, a
suitable approach could be first to apply genomic tiling arrays to identify novel tran-
scripts systematically under a wide range of experimental conditions, and then to use
the resulting improved annotation for the design of a comprehensive but cheaper
gene expression microarray.
As mentioned before, two-colour hybridizations are generally used for spotted
arrays because of variations in spot size and/or amount of probes. However, with
in situ
synthesized arrays, it can still be advantageous to perform competitive hybrid-
izations, in particular, using common reference designs that do not require special
analysis procedures and can be extended as long as the reference RNA is available.
The preferred choice for the common reference RNA is a pool of all samples
included in the study that provides an average expression profile. This requires
the gene ratios for the individual samples to reflect the expression levels relative
