Biological studies based on expression microarrays often depend crucially on comparison of expression measurements that are subject to technical variation associated with changes in time, laboratory or other measurement conditions. If a biological study involves different measurement conditions, then an experiment should be performed to gauge the associated technical variation and to compare it with the technical variation evident in replicate measurements made under the same measurement conditions. Such an experiment might be similar in design to the inter-laboratory study performed in the Microarray Quality Control (MAQC) project. This paper shows how one can infer, from MAQC-like data, the technical variation to be expected for materials of immediate biological interest. Using single-platform data from the MAQC project, we build a model of technical variation that includes assay-to-assay normalization, microarray response linearity, variance stabilization, variances of individual genes and covariances between pairs of genes, and site-to-site variation in microarray response.
Citation: Book chapter in
Pub Type: Others
expression microarrays, high dimensional measurements, inter-laboratory study, measurement error, normalization, statistical modeling