HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Full Text Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Google Scholar Articles by Han, E.-S. Articles by Hilsenbeck, S. G. Articles citing this Article PubMed PubMed Citation Articles by Han, E.-S. Articles by Hilsenbeck, S. G.

Reproducibility, Sources of Variability, Pooling, and Sample Size: Important Considerations for the Design of High-Density Oligonucleotide Array Experiments

¹ Department of Biological Science, The University of Tulsa, Oklahoma.
² Department of Physiology, The University of Texas Health Science Center, San Antonio.
³ South Texas Veterans Health Care System at San Antonio.
⁴ Breast Center at Baylor College of Medicine, Houston, Texas.

Address correspondence to Eun-Soo Han, Department of Biological Science, University of Tulsa, 600 S. College Ave., Tulsa, OK 74104. E-mail: eun-han{at}utulsa.edu

We have undertaken a series of experiments to examine several issues that directly affect design of gene expression studies using Affymetrix GeneChip arrays: probe-level analysis, need for technical replication, relative contribution of various sources of variability, and utility of pooling RNA from different samples. Probe-level data were analyzed by Affymetrix MAS 5.0, and three model-based methods, PM-MM and PM-only models by dChip, and the RMA model by Bioconductor, with the latter two providing the best performance. We found that replicate chips of the same RNA have limited value in reducing total variability, and for relatively highly expressed genes in this biologically homogeneous animal model of aging, about 11% of total variation is due to day effects and the remainder is approximately equally split between sample and residual sources. We also found that pooling samples is neither advantageous nor detrimental. Finally we suggest a strategy for sample size calculations using formulas appropriate when coefficients of variation are known, target effects are expressed as fold changes, and data can be assumed to be approximately lognormally distributed.