Figure 1

Flow-chart of the EvDTree method. 1- Initial data are pairs of superimposed protein structures. For each position, the "native" residue in the reference protein, the "substituted" residue in the superimposed protein and structural parameters are tabulated. 2- The data are filtered and grouped according to the "native" residue type, resulting in root clusters for classifications. 3- For each residue type, a hierarchical classification is achieved using a decision tree algorithm. Substitution scores are computed for all resulting clusters from residue distribution in the cluster.