Abstract
Clinical proteomic biomarker discovery experiments have produced few discoveries which have had an impact beyond their initial publication. In addition, there are far fewer validation experiments reported than discovery experiments. It has been suggested that one reason for this slow progress is a lack of application of standard concepts in experimental design. One such concept is the determination of sample size a priori to ensure that powerful investigations are undertaken, where there is a good chance of a difference being observed if one truly exists.
Proteomic experiments raise challenges in sample size determination due to their megavariate nature and inherent biological and technical variation. However, it is possible to determine sample size for investigations which seek to identify diagnostic, predictive and prognostic biomarkers. This can be undertaken by combining tools to control the expected false discovery rate, appropriate sample size formulae and simulation. A flexible protocol is described that allows the determination of sample size for a variety of proteomic biomarker discovery and validation experiments. These allow the determination of appropriate sample sizes to conduct suitably powerful experiments. It is possible that the application of such techniques could lead to more robust discoveries being validated and enter clinical practice, after appropriate evaluation.