Experimental design for large scale omic studies

Abstract

Molecular phenotyping has expanded from small sample sizes to larger complex studies and are now a common element in genetic studies. When large scale studies add a molecular phenotyping component, balancing omics batches for the factors of interest (e.g. treatment), regardless of the initial sample collection strategy always improves power. Where possible, confounding sources of experimental error that are not of interest (sample collection blocks and data collection plates) improves power as does planning batches for molecular phenotyping based on constraints during initial sample collection. Power for testing differences in molecular phenotypes is always higher when accounting for the entire experimental design during modeling. The inclusion of metadata that tracks sources variation is critical to our shared goals of enabling reproducible research.