Boruta Screening Feature Validation

[xuxwen]

In the use of Boruta screening variables with a dataset of screened features how to ensure the accuracy of the features obtained, do not need to go to the verification? For example, if I change a dataset with the same ending, maybe in a different population, re-screening, but do not get the original dataset screened features, that is, screened features do not intersect, so there is no way to illustrate the accuracy of the screened features, because there is no way to validate the screened predictors!