Fig. 1.
Diagram of dTERMen procedure. Target structure (A) is decomposed into TERMs guided by the graph of its coupled residues (B, Top) and the graph of residue-backbone influences (B, Bottom). Close matches to each TERM from the structural database are identified (note, higher-order TERMs were not considered in this study) (C), and the sequence alignments implied by these matches are used to estimate EPs governing the sequence–structure relationship in the target structure (D). Combinatorial optimization is then used to produce the optimal sequence for the target (E) or can also be used to build a library of design variants or for other tasks.