Joint modeling of phenotipic variables and gene expression data in early drug development experiments

Abstract

Drug development benefits enormously from a microarray experiment, a tool that allows accurate and relatively inexpensive collection of gene expression information for thousands of genes at a time. Recently, it has become a commonplace in biomedical research to monitor gene expression levels associated with different phenotypes. It is the aim of the investigator to determine which genes or combination of of genes could serve as biomarker for the $IC_{50}$. The joint modeling approach that allows the investigation of the relationship between the gene expression and the $IC_{50}$ after adjusting for the treatment effect was used in the selection and evaluation of genomic biomarkers. Depending on their intended use, biomarkers are further classified as prognostic and therapeutic. In the hope of achieving information gain, Supervised Principal Components Analysis (SPCA) was also conducted to construct a joint biomarker profile. Of the 7722 genes, 288 and 900 genes can serve as therapeutic