Omixer: a Bioconductor package for multivariate and reproducible sample randomization
Lucy Sinke, Davy Cats, and Bas Heijmans have developed a novel Bioconductor package for sample randomization in omics studies. It optimally distributes variables of interest across batches to proactively counter unwanted technical variation in the data.
Batch effects can overshadow biological differences in size and critically influence the results of omics studies. Even in benign cases, they decrease power to detect a true biological effect or contaminate results with false positives. Despite the numerous statistical methods developed to adjust for batch effects, a reactive approach is often insufficient. In fact, when technical variables are confounded with experimental factors of interest, batch effect correction will mask the underlying biological signal.
Sample randomization is a proactive, and arguably more impactful, method for obtaining reproducible results in high-throughput experiments. We developed Omixer—an R package for multivariate and reproducible randomization in omics studies. From a diverse range of randomized sample layouts, it selects the one that optimally balances biological variables across batches. Omixer offers the flexibility required to perform randomization effectively in a variety of study designs and experimental setups, and is available from Bioconductor.
Have a look here for more details: