Reply to Comment on: Tacrolimus in the treatment of childhood nephrotic syndrome: Machine learning detects novel biomarkers and predicts efficacy.

One of the purposes is to reduce the sample bias that may be caused by data imputation by controlling the number of missing variables and their missing rates. The machine learning algorithm is used to balanced samples to build a model, each variable can have at least 10 observations according to the common guideline.[1] Similarly, the same is true for the model construction in Group 2, and there are only 5 independent variables. [Extracted from the article]