Master regulator analysis workflow.

First, test cross-validation error for random forest models with increasing numbers of regulators to determine a parsimonious number of regulators sufficient to minimize prediction error. Next, identify the most important regulators as measured by the percent increase in the mean square error (MSE) upon permutation of the regulator in all trees of the forest. Finally, train the final random forest model with the determined number of top regulators by importance and test the model in the test set and validation set.

(PDF)