Statistical modeling in case-control real-time RT-PCR assays, for identification of differentially expressed genes in schizophrenia.

Aspects of experimental design, statistical modeling, and statistical inference in case-control real-time reverse transcription-polymerase chain reaction (RT-PCR) assays are discussed. The background is mRNA expression data from an investigation of genes previously suggested to be schizophrenia related. Real-time RT-PCR allows large samples of individuals. However, with more individuals than positions per plate, incomplete designs are required. A basic multivariate (for several genes jointly) random-effects analysis of covariance model, incorporating heterogeneity both between and within individuals, is formulated. The use of reference genes to form additional regressors is found to be highly efficient. Because regressions between and within individuals are usually different, it is important first to average over the intraindividual replicates. This has consequences for the influence of plate effects. Topics also discussed are testing for a significant mean disease effect, differential coregulation, and the difficulty of identifying genes affected in only a subgroup of cases.