alchemical free energy protocol with average network analysis
FESetup release 1.3dev, SUI version: 0.8.1, somd-freenrg 2016.1.1, pymbar version 3.0.0.dev0.dev-Unknown, Networkx v.1.11, networkanalysis v 0.1.1+23.ge21fdef
forcefield amber, ff14sb, tip3p, hfe
gaff2
mdengine, amber pmemd
AFE.type sire
AFE.separate_vdw_elec false
box.type = rectangular
box.length = 12.0
neutralize = False
temperature = 300 K
pressure = 1 atm
nmoves = 20000
ncycles = 100
buffered coordinates frequency = 5000
save coordinates = True
timestep = 2 * femtosecond
constraint = hbonds-notperturbed
hydrogen mass repartitioning factor = 1.0
cutoff type = cutoffperiodic
cutoff distance = 10*angstrom
barostat = True
andersen = True
energy frequency = 250
precision = mixed
minimise = True
equilibrate = False
equilibration iterations = 5000
center solute = True
reaction field dielectric = 82.0
minimal coordinate saving = True
lambda array varying between 9-27 evenly spaced lambda windows
The ligands were prepared for relative free energy calculations from the output of the docking protocols. Each ligand was parameterized using GAFF inside FESetup. Ligands were then solvated in TIP3P water and minimized and equilibrated for 100 ps. For the relative free energy perturbations, a series of alchemical transformation was proposed and the necessary perturbation protocol was computed with FESetup, resulting in Sire compatible input for the solvated ligands as well as vacuum ligands. FESetup, calling AMBERtoolswas used to solvated the protein and ligand complexes without adding any counter ions resulting in a charged simulation box. The solvated complexes were minimized and equilibrated for 100 ps with all but solvent restraint. Alchemical morphs are obtained with FESetup with explicit mapping of atoms for different binding poses of the same ligand. Production simulations were carried out using Sire somd-frenrg. Prior to the production run, Sire was also used for a further minimization and 2 ps equilibration to each of the intermediate lambda values using an annealing protocol. All simulations were run with a 2 fs timestep. Reduced gradients and energies were saved every 400 fs with a total simulation time of 4 ns. The simulation temperature of 298.15 K was kept constant using an anderson thermostat and the pressure of 1 atm was maintained using a MC barostat as implemented in OpenMM. The barostat frequency is set to 25 and the collision frequency of the anderson thermostat is 1 /ps. Furthermore, multi state Bennet's acceptance ratio (MBAR) was also as a second free energy analysis method. All submitted results come from the MBAR estimator including the error estimates. 5% of the initial data was discarded to equilibration in the MBAR analysis. The individually estimated free energy differences were then read into a networkx digraph. Backward and forward simulations were averaged, as well as repeated runs. All possible paths within the graph were estimated and weighted averages of all possible paths between two compounds for the relative free energy computation were computed. Standard errors were computed based on forward and backward averages as well as repeated runs and then propagated along the paths using standard error propagations. Each path was then weighted based on the standard error along the path, the same was done for averaged path errors. Due to large range in potencies based on these calculations a set of scaling calculations were run, slowly scaling the partial charges generated by gaff2 from 100% of the original charge to 70% charge for all neutral compounds and to 50% of the original partial charges for charged compounds. This was done using 5 lambda windows for each compound bound to the protein and in the solvated phase. The estimated charging corrections for each compound were then added to the free energy estimates from the path averages all with respect to the chosen reference compound.
Maestro/ Released crystal structure
Maestro 11 (Beta Version)
Assumed pH 7.4
Tautomers considered
Released structure 1ymrc was used as initial structure to build the receptor. The missing fragment Y264 - K279 was modelled from the FXR-FXR_17 complex structure 1hqmf released by the organisers. Sidechains within 4 Angstrom of FXR17 were rearranged in the final receptor structure to resemble 1hqmf orientation.