Reducing False Positive Findings in Statistical Analysis of Pharmacogenomic Biomarker Studies Using High-Throughput Technologies

The promise of pharmacogenomics lies in the potential to establish a personalized drug therapy with the intent of maximizing effectiveness and minimizing risk, through development of pharmacogenomics biomarkers. However, currently, most pharmacogenomic measurements are not considered valid biomarkers with clear clinical significance, thus this field is in early developmental stages. Recently, the development of comprehensive, high-throughput technologies such as gene expression microarrays has provided powerful new tools for these stages. This technological transformation is, at the same time, generating an increasing demand for statistical analysis of large and complex multivariate datasets from high-throughput assays. This article provides a review of the key features to be observed in statistical analyses of large amounts of data from pharmacogenomic biomarker studies with high-throughput assays. The problem of false positive can be very serious in such studies. The evaluation of stability and reproducibility of the results of statistical analysis are claimed to reduce chance that false positive findings are subject to further investigation in subsequent studies.