Get access

Biplot methodology in exploratory analysis of microarray data



Although principal component analysis is widely used in the exploration of microarray data, the advantages of constructing a biplot as multivariate analog to a scatterplot is seldom exploited. This paper illustrates the benefits of using biplots with microarray data to (1) visually display both the treatments and genes of such extreme high-dimensional data in a single plot, (2) relate the treatments to the underlying biological process through the use of biplot axes, and (3) to optimally separate classes and explore the differentially associated expression in genes. In this analysis, we have used gene expression measurements from human bronchial epithelial cells following exposure to whole cigarette smoke. Specifically, when exploring differences between treatments and differentially expressed genes, it is shown why the principal component biplot is not optimal and the analysis of distance biplot is introduced as an alternative to principal components. Copyright © 2009 Wiley Periodicals, Inc., Statistical Analysis and Data Mining 2: 135-145, 2009