Design and analysis of two-phase experiments for gene expression microarray experiments

Abstract

Gene expression microarrayexp eriments are intrinsicallyt wo-phase experiments. Messenger RNA (mRNA), required for the microarrayexp eriment, must first be derived from plants or animals that are exposed to a set of treatments in a previous experiment (Phase 1). The mRNA is then used in the subsequent laboratory-based microarray experiment (Phase 2) from which gene expression is measured and ultimatelyanaly zed. We show that obtaining a valid test for the effects of treatments on gene expression depends on the design of both the Phase 1 and Phase 2 experiments. Examples show that the multiple dyeswap design at Phase 2 is more robust than the alternating loop design in the absence of prior knowledge of the relative size of variation in the Phase 1 and Phase 2 experiments.