We introduce the GeWare data warehouse approach for microarray-based gene expression analysis. GeWare centrally stores raw and preprocessed expression data together with a variety of annotations to support different analysis forms. The flexibility of GeWare is made possible by a multidimensional data model where expression data is stored in several fact tables which are associated with multiple hierarchical dimensions holding describing metadata on genes, samples, experiments, and processing methods. Annotations are uniformly managed in a generic way thus supporting the high variability of annotations and easy extensibility. Gene annotations are imported from several pub-lic sources and integrated with each other based on commonly used accession numbers, e.g. Uni-Gene clusters.