Figure 2.

Variable importance measures for the combined dataset. Numbered positions represent nucleotide state variables (with position zero representing the edited/non-edited site). The importance of each position is the decrease in the Gini index (a measure of impurity) induced by splitting the data on that position averaged over all trees (higher values are more important). The three variables based on estimates of free energy of folding are the codon position of the edited site (cp), estimated free energy of folding for the entire 41-nucleotide sequence centered on the edited/non-edited site (fe), and the difference in estimated free energy of folding between the edited and non-edited versions of the 41-nucleotide sequence (dfe).