Figure 3

Outline of the basic analysis workflow of A-MADMAN. Once data and metadata are downloaded, organized and annotated, tag-based queries can be used to create data baskets. In the figure two data baskets are represented, each including samples obtained from three platforms (circles represent samples; red and blue borders represent two different baskets, corresponding for instance to healthy and diseased individuals; yellow, green ad light blue filling represent different platforms; numbers indicates specific individuals; squares stand for single expression values). RMA is carried out separately for each platform, according to the selected CDF, and gene expression levels and metadata, originating from user-defined annotations, are incorporated into a number of ExpressionSet objects corresponding to represented platforms. The annotations are used to generate a non-redundant list of gene-centered identifiers, which become the rows of a data matrix filled with integrated expression values, obtained averaging the expression levels of the various platform-specific probesets/metaprobesets (as shown for expression level of gene X in individual 1). After a final normalization step, by quantile distribution transformation, the integrated ExpressionSet object is produced.