Skip to main content
Premium Trial:

Request an Annual Quote

Readers Tips, How-To's, Experiences: Nov 1, 2004


Q: What’s your top advice for designing a microarray analysis experiment to ensure that your statistical analysis, control for error, and normalization procedures will give you an accurate end answer?


A. I am going to enclose a couple of papers by Richard Simon at the NCI who is the expert in the questions you have asked. To be very honest, a large proportion of array experiments in the past were of very poor quality and no amount of clever statistical manipulation would give you good results (the [garbage in = garbage out] principle). The array companies are putting out very good slides and our “roll your own” are much more reliable now because of better printing instruments, substrates (slides), etc.

Journal of the National Cancer Institute, Vol. 95, No. 18, September 17, 2003, pp. 1362-1369

BMC Bioinformatics 2003, 4:33, 02 September 2003

Also available through http://www.

Ernest Kawasaki
Advanced Technology Center
National Cancer Institute, NIH


A. My number one piece of advice is to be absolutely sure that all samples are prepared using the same method. If a subset of samples are limiting, then do all samples using the protocol for the most limiting method. Each sample preparation method has an inherent rate of reproducibility, but it will also have an inherent bias. This bias is typically consistent for all samples prepared by that method. If you use two different methods you will be adding a new, uncontrolled factor to your study which will likely make it more difficult to analyze your results.

My second piece of advice is don’t limit yourself on the number of experimental replicates. You will get better data doing one 6x6 comparison than by doing 3x3x3x3. Let the power of your study work for you!

Chris Barker
Director, Genomics Core Laboratory
Gladstone Institute of Cardiovascular Disease
University of California, San Francisco


A. I’d say that the most important things are sufficient replication, preferably at the level of biological replicates.

Understand your sources of non-biological variation, and try and design it such that those sources will not confound the biological variation. That is, if you are running 20 arrays, 10 from treated, and 10 from untreated samples, don''''t do all the treated on one day, and the untreated on another. The same idea holds for any potential source of variation, e.g. if you’re using different microarray batches, or different hybridization chambers or water baths.

Gavin Sherlock
Director, Microarray Informatics
Stanford University


A. Most researchers use reference designs when doing two-color cDNA microarray experiments and only print one spot per gene. My advice to researchers using two-color arrays is to print the slides with several replicate spots per slide to provide insurance against bad spots and improve the control of error at low cost. Secondly, to use balanced, incomplete block designs (i.e. loop designs) whenever possible to reduce the cost and increase the accuracy of the results for the same number of slides. Finally, discuss the experiment with a statistician to insure that the true replication of the experiment is being done at the level of the independent biological samples.

James L Rosenberger
Department of Statistics
Penn State University


A. I’d like to emphasize a few key points special for two-color microarray format:

1) Confirm informatics of the array design and layout to be correct.

2) Use as few treatment conditions as possible.

3) Sample RNA from separate biological resources for a replicated microarray experiment.

4) Use high-quality RNA.

5) Use a common reference RNA for each hybridization.

6) Use equal length of cDNA for Cy3- and Cy5-labeled probes.

7) Use equal amount of dye concentrations for the labeled probes.

8) Use exogenous nucleic acid (external RNA) as spiking-in controls.

9) Use the control genes as a reference for the first step of normalization and estimate CV for the microarray.

Z. Lewis Liu
Research Molecular Biologist
National Center for Agricultural Utilization Research


The Scan

Lung Cancer Response to Checkpoint Inhibitors Reflected in Circulating Tumor DNA

In non-small cell lung cancer patients, researchers find in JCO Precision Oncology that survival benefits after immune checkpoint blockade coincide with a dip in ctDNA levels.

Study Reviews Family, Provider Responses to Rapid Whole-Genome Sequencing Follow-up

Investigators identified in the European Journal of Human Genetics variable follow-up practices after rapid whole-genome sequencing.

BMI-Related Variants Show Age-Related Stability in UK Biobank Participants

Researchers followed body mass index variant stability with genomic structural equation modeling and genome-wide association studies of 40- to 72-year olds in PLOS Genetics.

Genome Sequences Reveal Range Mutations in Induced Pluripotent Stem Cells

Researchers in Nature Genetics detect somatic mutation variation across iPSCs generated from blood or skin fibroblast cell sources, along with selection for BCOR gene mutations.