Research article

Incorporation of gene-specific variability improves expression analysis using high-density DNA microarrays

Vikram Budhraja^1,2, Edward Spitznagel², W Timothy Schaiff¹ and Yoel Sadovsky^1*

• * Corresponding author: Yoel Sadovsky sadovskyy@wustl.edu

Author Affiliations

¹ Department of Obstetrics and Gynecology, and Cell Biology and Physiology, Washington University School of Medicine, St. Louis, MO 63110, USA

² Department of Mathematics, Washington University, St. Louis, MO 63130, USA

For all author emails, please log on.

BMC Biology 2003, 1:1 doi:10.1186/1741-7007-1-1

Published: 28 November 2003

Abstract

Background

The assessment of data reproducibility is essential for application of microarray technology to exploration of biological pathways and disease states. Technical variability in data analysis largely depends on signal intensity. Within that context, the reproducibility of individual probe sets has not been hitherto addressed.

Results

We used an extraordinarily large replicate data set derived from human placental trophoblast to analyze probe-specific contribution to variability of gene expression. We found that signal variability, in addition to being signal-intensity dependant, is probe set-specific. Importantly, we developed a novel method to quantify the contribution of this probe set-specific variability. Furthermore, we devised a formula that incorporates a priori-computed, replicate-based information on probe set- and intensity-specific variability in determination of expression changes even without technical replicates.

Conclusion

The strategy of incorporating probe set-specific variability is superior to analysis based on arbitrary fold-change thresholds. We recommend its incorporation to any computation of gene expression changes using high-density DNA microarrays. A Java application implementing our T-score is available at http://www.sadovsky.wustl.edu/tscore.html webcite.