Identifies an optimal transformation of a surrogate marker such that the proportion of treatment effect explained can be inferred based on the transformation of the surrogate and nonparametrically estimates two model-free quantities of this proportion. Details are described in Wang et al (2020)