Skip to main content
. 2007 Jan 5;92(6):2120–2130. doi: 10.1529/biophysj.106.096016

FIGURE 2.

FIGURE 2

(A) Local region of the protein (1–5 consecutive residues around the targeted ligand) is randomly chosen for perturbation. The ϕ and ψ angles of the targeted position (as shown in the circle) are perturbed by up to 5°. (B) All amino acid rotamers consistent with these torsion angles are selected at each position from the Dunbrack and Cohen rotamer library (35,36). Rotamer-backbone and rotamer-rotamer energies are calculated for all the selected rotamers. (C) The binding energy is minimized using a MILP formulation to select the optimal rotamer at each of these positions. (D) After rotamer selection for the target molecule, the nontarget ligand is docked, and its binding energy is calculated using the CHARMM energy function (18,19). The best conformation (i.e., rotamer choices) is accepted if it ensures that the binding score for the undesirable ligand(s) remains greater than what is needed for successful binding. When this threshold is exceeded, the corresponding design choice is deemed infeasible. (E) In this step the backbone and the targeted ligand are allowed to relax to adjust to the changes in the side chains. This is achieved by allowing ϕ and ψ to vary freely and to be determined during energy minimization. (F) ZDOCK software is employed to readjust the targeted ligand position regarding the modified backbone and side chains (20). (G) Protein-ligand binding energy is computed using CHARMM energy functions. If the binding energy for the target ligand is lower than the previous best ligand structure then this move is accepted as best solution otherwise. (H) Metropolis criterion is used to decide whether to accept or reject the move.