Modeling of bioassay data and genes expression in drug discovery experiments: A supervised principal component analysis approach

Abstract

Nowadays, microarray technology is extensively used in biological and medical studies to monitor simultaneously the activity of thousands genes and their response to a certain treatment. In some experiment, in addition to gene expression, some responses are available and the question of interest is to identify whether or not the gene expressions can serve as biomarkers, and perhaps, as surrogate for the response. However, it is challenging to analyze the microarray data due to high dimensionality of the data and relatively small number of observations. Moreover, although the number of genes assayed is large, there may be only a small number of biomarkers that are associated with variations of the outcome. Thus, selection of genes, for which gene expression might serve as biomarker for the outcome, is needed. In our study supervised principal component analysis has been used to construct a joint biomarker. A subset of genes that are associated with the outcome is selected, and is used to build the joint biomarker.