Combining not-proper ROC curves and hierarchical clustering to detect differentially expressed genes in microarray experiments

Abstract

TNRC (Test for Not Proper ROC Curve) is a statistical tool recently developed to identify differently expressed genes in microarray studies. In previous investigations it was demonstrated to be able to separate hidden subgroups in a two-class experiment, but being a univariate technique it could not exploit the complex multivariate correlation naturally occurring in gene expression data. In this study we show as the combination of TNRC with a standard technique of hierarchical clustering may provide useful biological insights. An example is provided using data from a publicly available data set of 4026 gene expression profiles in 42 samples of lymphomas and 14 samples of normal B cells. © 2014 Springer International Publishing Switzerland.