Study Demonstrates Gene Expression Microarrays are Comparable and Reproducible
A study funded by the National Cancer Institute, part of the National
Institutes of Health, shows for the first time that microarray
data generated in different laboratories can produce highly comparable
results. For this comparison study, appearing in the Jan. 15, 2005,
Clinical Cancer Research*, four separate laboratories analyzed
gene expression (whether genes are turned on or off) for the same
set of human tumor tissues. Overall, the expression profiles of
portions of individual samples were highly comparable, and the
experimental correlation between separate labs was only slightly
lower than correlation of duplicated experiments within the same
“This study is a key first step in moving gene expression data from
small-scale bench science into large-scale clinical evaluation,” said
James Jacobson, Ph.D., chief of NCI’s Diagnostic Biomarkers
and Technology Branch.
Gene expression microarrays have been used in numerous applications,
including identifying novel genes associated with certain cancers,
classifying tumors, and predicting patient outcome. So far, though,
microarray studies have been performed by individual institutions.
Evaluation of the potential clinical use of microarrays may require
larger studies carried out in multiple locations and would necessitate
that microarray data produced in different laboratories be combined
for analysis. Even if all procedures and equipment were the same,
small differences between labs, such as in handling the tissue
samples, extracting the RNA, or scanning the microarrays, could
result in different profiles.
To test the feasibility of multi-lab microarray studies, the NCI’s
Director’s Challenge program set up this preliminary study
to compare results between labs. Twelve different tumor tissues,
five cancer cell lines, and five purified RNA samples were prepared,
blinded, and randomized for analysis in four laboratories: University
of Michigan Medical School, Ann Arbor, Mich.; Dana Farber Cancer
Institute/Whitehead Institute, Boston, Mass.; Memorial Sloan-Kettering
Cancer Center, New York, N.Y.; and H. Lee Moffitt Cancer Center,
Tampa, Fla. All five cell lines and RNA samples were derived from
lung carcinomas, while the tissue samples were derived from multiple
normal and cancerous tissues. The labs followed a common protocol
for all steps of sample preparation and microarray analysis and
the resulting gene expression profiles were analyzed and compared.
Sample correlation within each lab was extremely high; microarray
data from the RNA samples had the highest correlation values, followed
by the cell lines, and then finally the tissue samples. This was
not surprising, as RNA required the fewest steps of preparation,
while tissues required the most. The between-lab correlation values
decreased slightly for all samples, but in all cases expression
profiles of similar samples could be accurately grouped together.
The researchers also examined variations of individual gene measurements
to help determine causes of variability, and they found that laboratory
practices comprised the smallest source of variation in these studies.
Measurement errors common to all labs were the next most common
contributor, and biological differences in different samples were
the largest source of variation.
“This study indicates that microarrays have a high degree of reproducibility,
as long as standardized protocols are carefully followed,” said
Jacobson. These promising results will allow this same research
group to proceed with a larger gene expression analysis of 600
stage I lung adenocarcinomas, with the hopes of confirming a previous
association between gene expression profiles and patient outcome.
This project is also an example of NCI interest in developing public/private
partnerships. Affymetrix, of Santa Clara, Calif., contributed part
of the arrays for this comparison study and provided technical assistance to the four sites carrying out the
study. Ardais Corporation, Lexington, Mass., provided the RNA samples
used for analysis.
For more information about the Director’s Challenge program,
please visit http://dc.nci.nih.gov/.
For more information about cancer, please visit the NCI Web site
at http://www.cancer.gov or call NCI's Cancer Information Service
at 1-800-4-CANCER (1-800-422-6237).
* Kevin K. Dobbin, David G. Beer, Matthew Meyerson, Timothy J. Yeatman, William L. Gerald, James W. Jacobson, Barbara Conley, Kenneth H. Buetow, Mervi Heiskanen, Richard M. Simon, John D. Minna, Luc Girard, David E. Misek, Jeremy M.G. Taylor, Samir Hanash, Katsuhiko Naoki, D. Neil Hayes, Christine Ladd-Acosta, Stecen A. Enkemann, Agnes Viale, and Thomas J. Giordano. Interlaboratory Comparability Study of Cancer Gene Expression Analysis Using Oligonucleotide Microarrays. Clinical Cancer Research, Vol. 11; Jan 15, 2005.