Principles for the Design & Analysis of Experiments Using Gene Expression Arrays

High-throughput assay methods such as gene expression arrays, proteomics by mass spectrometry, and metabolomics by NMR spectroscopy present many data analytical challenges. In this talk, I present a coherent strategy for the important task of determining differential gene expression (or differences in protein and metabolite concentrations). This strategy is based on a combination of standard statistical methods such as regression and the analysis of variance, methods for accounting for multiple testing such as false discovery rate methods, and innovations in data transformation to make these methods more effective. Software to implement these methods is available.