Performance of the KIC loop reconstruction protocol. (a) Representative set of 12-residue loop reconstructions (blue) on dataset 2. PDB identifiers and r.m.s. deviation to the crystallographic loop (cyan) are shown. (b) Box-plot comparison of the standard Rosetta and KIC Rosetta protocols on dataset 1 (left), both Rosetta protocols with the molecular mechanics method on dataset 2 (middle), and the KIC Rosetta protocol on dataset 3 (right). Boxes span the interquartile range (IQR, 25th-75th percentiles), black lines represent the median, whiskers extend to furthest values within 0.8 times the IQR, and open circles are outliers. (c) KIC reconstruction of conformational changes in the Rac switch I loop when bound to ExoS toxin (blue reconstruction on cyan crystal structure, blue partner; PDB 1he1) or Rho guanine dissociation inhibitor (orange reconstruction on purple crystal structure, orange partner; PDB 1hh4).

Loop reconstruction with KIC. (a) In the KIC move, 3 Cα atoms of an N-residue chain are designated as pivots (green spheres); the remaining N – 3 are non-pivot Cα atoms (cyan spheres; left). In a 12-residue loop, 24 torsions are modeled. Nonpivot torsions are sampled from a residue type-specific Ramachandran map, opening the chain (middle). KIC then finds all values for the pivot torsions that close the loop, if any exist, keeping the endpoints fixed (right). The previous state is shown in outline. (b) Performance of the Rosetta KIC protocol and standard protocols on a 12-residue loop (Protein Data Bank (PDB): 1srp). Only KIC densely sampled regions < 1.0 Å r.m.s. deviation from the crystallographic loop. Asterisks mark the lowest-scoring reconstructions from the two methods. The Rosetta all-atom score includes the enthalpy plus the solvation contribution to the entropy but not the configurational entropy. (c) The lowest scoring reconstructions from b are shown. KIC improved reconstruction accuracy to 0.6 Å from 2.6 Å using the standard protocol.