Design issues in toxicogenomics using DNA microarray experiment.

The methods of toxicogenomics might be classified into omics study (e.g., genomics, proteomics, and metabolomics) and population study focusing on risk assessment and gene-environment interaction. In omics study, microarray is the most popular approach. Genes falling into several categories (e.g., xenobiotics metabolism, cell cycle control, DNA repair etc.) can be selected up to 20,000 according to a priori hypothesis. The appropriate type of samples and species should be selected in advance. Multiple doses and varied exposure durations are suggested to identify those genes clearly linked to toxic response. Microarray experiments can be affected by numerous nuisance variables including experimental designs, sample extraction, type of scanners, etc. The number of slides might be determined from the magnitude and variance of expression change, false-positive rate, and desired power. Instead, pooling samples is an alternative. Online databases on chemicals with known exposure-disease outcomes and genetic information can aid the interpretation of the normalized results. Gene function can be inferred from microarray data analyzed by bioinformatics methods such as cluster analysis. The population study often adopts hospital-based or nested case-control design. Biases in subject selection and exposure assessment should be minimized, and confounding bias should also be controlled for in stratified or multiple regression analysis. Optimal sample sizes are dependent on the statistical test for gene-to-environment or gene-to-gene interaction. The design issues addressed in this mini-review are crucial in conducting toxicogenomics study. In addition, integrative approach of exposure assessment, epidemiology, and clinical trial is required.