Fig. 2: Relative size of feature categories and feature similarity distributions.
From: Machine learning differentiates enzymatic and non-enzymatic metals in proteins
a Distribution of features used for training. The four groups of Rosetta terms each include 84 features calculated in one of four ways—the mean or average of residues within four shells or spheres—for a total of 294 unique Rosetta category features since the first shell and sphere are the same. b, c The kernel density estimations of feature similarity between enzymatic and non-enzymatic sites for each Rosetta calculation method (b) and the other four feature categories (c).
