A variable fold change threshold determines significance for expression microarrays.

The use of expression microarrays to determine bona fide changes in gene expression between experimental paradigms is confounded by noise due to variability in measurement. To assess the variability associated with transcript hybridization to commercial oligonucleotide-based microarrays, we generated a data set consisting of five replicate hybridizations of a single labeled cRNA target from three distinct experimental paradigms, using the Affymetrix human U95 GeneChip set. We found that the variability of expression level in our data set is intensity-specific. We quantified the observed variability in our data set in order to determine significant changes in gene expression. LOESS fitting to a plot of the standard deviation of replicates assigned a variability associated with a specific intensity. This allowed for the calculation of a "variable fold-change" threshold for any absolute intensity at any level of statistical confidence. Testing of this method indicates that it removes intensity-specific bias and results in a 5- to 10-fold reduction in the number of false-positive changes. We suggest that this approach can be widely used to improve prediction of significant changes in gene expression for oligonucleotide-based microarray experiments and reduce false leads, even in the absence of replicates.