Computational protocol: In Silico Prediction of Mutant HIV-1 Proteases Cleaving a Target Sequence

Similar protocols

Protocol publication

[…] The structure of wild type (WT) HIV-1 protease in complex with different octa-peptides was optimized using PyRosetta 1.1 , a python script-based interface to Rosetta , and the algorithm depicted in . The algorithm is based on the flexible peptide-docking algorithm used by Chaudhury and Gray to identify in WT HIV-1 protease the active-site residues mostly involved in the discrimination of cleavable and non-cleavable peptides. Following their algorithm, the HIV-1 protease – peptide complexes are represented in atomic resolution, as opposed to a coarse-grain representation. With respect to the algorithm described in , our algorithm () has a larger number of cycles (8×4×6 = 192 compared to 8×12 = 96), and more ‘small’ and ‘shear’ moves for the perturbation of both the side chain and the backbone atoms. The side chain conformations are further optimized through a repacking algorithm and using the extended Dunbrack library , . The moves are applied to all residues of the substrate peptide plus a selected number of residues of the protease, with the following criterion: all residues inside a 5 Å distance from any atom of the substrate peptide, plus all the residues reported as active by Chaudhury and Gray , plus their 1 neighbours, plus if one residue is included on only one chain it is made to be included in both. After the moves, an energy minimization step is performed, based on the Davidon-Fletcher-Powell method , . Each structure is then accepted or rejected based on a Monte Carlo (MC) criterion depending on the standard RosettaDock energy function , , –. Along the optimization a temperature gradient was applied, from an initial value of kT = 3.0 to 1.0, unless differently stated. 500 decoy structures were generated using 5 parallel algorithm runs, each producing 100 structures.The main difference with the algorithm of is that after the algorithm produced 500 decoy structures, the lowest in energy is chosen and used as a starting structure for another cycle of optimization. This process is repeated times, until convergence. It was found that, after at least 5 cycles, the computed RosettaDock energy did not change between subsequent cycles as soon as all 5 parallel runs of a single cycle produced structures with the same energy. Consequently, in order to render as automatic as possible the algorithm, the fact that and that each parallel run produced, as best structure, a decoy with the same energy was taken as a mark for convergence. It was found that, on average, a value of was sufficient. As an example, reports the energy of WT-PR bound to TF-PR along the optimization. The points at each step corresponds to the RosettaDock energy of the lowest in energy decoy out of the 500 computed at that particular step. Such structure would then be used as starting point for the next cycle. At the end of the cycles the lowest in energy decoy is chosen as the PyRosetta optimized structure.The same algorithm was also used for the optimization of mutant HIV-1 proteases (vide infra), the octa-peptides alone, and the protease alone as apo-protein.The starting structures were prepared from that of HIV-1 protease in complex with an inhibitor (PDB accession code 1HXB ), considered as apo-protein. In order to place the substrate peptide, the structure of a D25N deactivated protease in complex with the natural substrate peptide p2-NC (PDB accession code 1KJ7 ) was aligned with respect to the backbone atoms of the protease (RMS = 0.436 Å). The starting structure was then composed using the apo-protein from 1HXB and the substrate peptide from 1KJ7. All subsequent protease-peptide complexes were created starting from this structure and mutating the peptide accordingly. See for a complete list of the considered substrate peptides. Hydrogen atoms were added through the program Pymol . [...] The position of the hydrogen atoms of each PyRosetta generated structure was optimized using Open Babel with the MMFF94 – force field. The energy of each structure was finally re-evaluated at the higher level of theory ‘FMO2-MP2/6-31G(d)/PCM ’. Single point energy evaluations were carried out using the fragment molecular orbital (FMO) approximation , , as implemented in GAMESS . Each FMO calculation was carried out at the MP2 level of theory with the 6–31 G(d) basis set , and the Polarazible Continuum Model (PCM) approximation , . Pairs of fragments separated by more than two van der Waals radii were calculated using a Coulomb expression for the interaction energy and ignoring correlation effects (RESDIM = 2.0 RCORSD = 2.0 in $FMO). The input files for the FMO calculations were prepared using the program FRAGIT . […]

Pipeline specifications

Software tools PyRosetta, RosettaDock, PyMOL, Open Babel
Applications Drug design, Protein interaction analysis
Organisms Human immunodeficiency virus 1
Diseases HIV Infections