Genomincs &Bioinformatics Research &Support Program of

Diabetes, Digestive, Kidney Diseases

Investigators

Linked publications & trials

Abstract

Over recent years, a variety of microarray platforms, both commercial and custom-made, have become increasingly available for use in genome-wide gene expression studies. Subsequently, a large number of studies have incorporated the use of multiple array platforms to determine the compatibility of data derived from different sources. Comparisons of the data generated from the different array formats have yielded mixed results, with some finding overall satisfactory to good correlations, while others finding them to be poor. Because of the wide range of experimental designs, array types, and statistical and filtering approaches that support the conclusions of these studies, it is difficult to come to a comprehensive statement on the overall agreement of microarray gene expression data across multiple platforms. The underlying cause of diversity in the observations regarding cross-platform comparability probably arise from several factors, including the magnitude of gene expression differences across samples being assayed, the use of biological and/or technical replicates and associated variability measures therein, the use of signal thresholds to limit ?noise? in the measurements, and the accuracy of public gene identifiers used to annotate and combine probes from multiple platforms. We have found that gene expression changes obtained from biological samples are often discordant across platforms, using combined fold-change and other statistical criteria. This means that when data are collectively analyzed across two or more platforms, that the total number of genes found to be differentially expressed is greater than if the experiment been run on a single platform, given a common set of genes. Having performed real-time PCR validation on a subset of genes that were found differentially expressed either by one, two or three platforms, we have found that when change is detected by two or more platforms, that the likelihood for this change to be true is high, if not absolute, which allows one to validate gene expression changes measured at a genome-wide level using alternate array platforms.