It estimates the protein abundance variance from the fitted model and simulates data with certain number of biological replicates based on the variance estimation. It reports the mean predictive accuracy of the classifier and mean protein importance over multiple iterations of the simulation. While varying the number of biological replicates to simulate, the sample size which generates the largest predictive accuracy is estimated. And the proteins which can best separate different conditions are reported.

The package also uses the fitted models and the fold changes estimated from the models to calculate sample size for hypothesis testing. It outputs the minimal number of biological replicates per condition to acquire the expected FDR and power under different fold changes.

The R-Shiny tool provides a fast and easy interface to exploit the MSstatsSampleSize package conduct data simulation tests and try out multiple simulations parameters for estimating an optimal sample size. The app is completely self contained and hosted on the Shiny-Server for open use. High dimensional data can get ingested by the app and can be deployed locally or on private h2o clusters. The source code available freely on github for custom interpretations and use cases