5Microarrays—Planning Your Experiment
Jean Yee Hwa Yang
Summary
The rapid increase in the use of microarray studies has generated many questions on how
to plan and design experiments that will effectively utilize this technology. Investigators often
require answers to questions relating to microarray platforms, RNA samples, options for repli-
cation, allocation of samples to arrays, sample sizes, appropriate downstream analysis, and many
others. Careful consideration of these issues is critical to ensure the efficiency and reliability of
the actual microarray experiments, and will assist in enhancing interpretability of the experimental
results.
Key Words: Experimental design; microarray; gene expression; probe design; replication;
randomization.
1. Introduction
Good experimental design in microarray studies simplifies analysis and
enhances interpretation of data. Various considerations go into the planning
of an effective experiment. In the last few years, many of the publications on
experimental strategies have focused on the identification of an efficient design.
While this is still an important component, there are broader considerations. In
this chapter, the various aspects involved in planning a successful microarray
experiment will be described.
From: Methods in Molecular Medicine, Vol. 141: Clinical Bioinformatics
Edited by: R. J. A. Trent © Humana Press, Totowa, NJ
71

72 Yang
2. Materials
When planning an experiment, a number of general issues need to be
identified to help translate the overall biological questions into a more defined
and appropriate statistical framework. Some of these include
1. Aim of the experiments: This refers to the biological question of interest. Given
that microarrays are utilized in a wide variety of contexts, this consideration can be
very specific or general. For example, a researcher can study a focused hypothesis
such as identifying differences between wild-type and mutant mice. Alternatively,
a research can perform an experiment with a very general aim in mind, such as
profiling gene expression from a collection of clinical patients and then attempt
to generate a possible hypothesis from the data. Regardless of the type of aim, it
is important to consider how these aims will contribute to understanding further
the long-term goal of the research (see Note 1 and Chapters 6–9 for examples of
different microarray studies including experimental designs).
2. Main comparison: Researchers often wish to investigate multiple questions in a
single experiment. Given that each question has implications for experimental design
and downstream analysis, researchers need to clarify the specific questions being
asked and subsequently, the most important question or comparisons. One such
example is the intention to derive simultaneously a classification rule that best
predicts survival outcome of cancer patients as well as to identify a collection of
biomarkers that best distinguish the survival outcome.
3. Resources: This determines the size and scale of the experiment and is predomi-
nantly dependent on the number of samples available, the amount of mRNA and
ultimately funding for the experiment.
4. Previous experiments: It is not always clear whether changes between mRNA are
detectable by microarray experiments with many of the genome-wide profiling
studies. For such situations pilot studies to determine the feasibility of a larger study
should be considered.
5. Verification method: It is possible to overlook the limitation of the arrays and the
way in which these data contribute to the overall research goal as researchers get
caught up in the novelty of a technology. As microarray is still a relatively new
experimental approach, large variability in results can be expected. Consideration
needs to be given to methods required to verify results obtained from microarrays.
3. Methods
There are two main aspects to planning a successful microarray experiment.
The first concerns the actual design or selection of the microarray. This refers
to the choice of DNA probes to print onto the solid substrate, e.g., a membrane,
glass slide, or silicon chip, and where they are to be printed. The second aspect
concerns the planning and design of the actual hybridizations. This often refers
to the allocation of target samples to the microarrays, the nature, and the number
of replications required (see Note 2).