MMGBSA Pose Prediction from 20 replicates of molecular dynamics simulations
Assumed pH 7.4
1. Starting structure:Ligands were prepared using Schrödinger’s standard ligand preparation at pH 7.4. The stereochemistries were given in the isosmiles strings and were therefore kept. Proteins were prepared using Schrödinger’s protein preparation wizard. Bond orders were assigned, hydrogen atoms added, disulfide bonds were generated, and water orientations were sampled. The charge states were kept at pH 7.0. Missing loops and side chains were constructed. The given residue confirmations were kept, with no alternative residue conformations examined.
The best ranked ligand docking structures using Schrödinger’s Glide or CCDC Gold were submitted for molecular dynamics simulations using AMBER14.
MD simulation: Use AMBER14 to equilibrate the protein-ligand complex starting structure for 5 ns, then run additional 20 replicates with 5ns each by restarting from the identical equilibrated atomic coordinates but different initial velocities generated from a Maxwell-Boltzmann distribution. The last 1.5 ns of each replicate was used as the production, which generated 300 snapshots in total. AMBER MMGBSA method was used to analyze the free energy. Entropy contribution was not included.
use MD simulation to predict the binding pose.
Rank 1: Cluster the 300 snapshots by structural similarity, and use the average structure from the highest population.
Rank 2: Select 30 lowest dGtot snapshots out of 300, and use the average structure of the selected conformations.
Rank 3: Select 30 lowest dGvdw snapshots out of 300, and use the average structure of the selected conformations.
Rank 4: Select 30 lowest dGelec snapshots out of 300, and use the average structure of the selected conformations.
Rank 5: Use the structure with the lowest dGtot from the 300 snapshots.