Rank multiple putative poses

Introduction

The rescoring simulation aims to rank binding poses. In order to achieve this, we developed a curated side chain prediction and minimization algorithm to relax the system. This simulation is generally used you need to discriminate between multiple plausible poses.

Inputs

  • protein-ligand PDB file(s)

  • YAML file with parameters

Default parameters

Recommendations

  1. Expected computational time: 2h

  2. We recommend using at least 50 CPUs.

1. Complex Preparation

Prepare the system consisting of protein and docked ligand with Schrödinger Protein Preparation Wizard. We would usually recommend protonating the protein (obligatory), deleting water molecules more than 5Å away from ligands and ions as well as filling in missing loops and side chains.

Make sure the ligand has:

  • unique chain ID

  • unique PDB atom names with no spaces or single letters

  • any residue name except for UNK

2. Input Preparation

Prepare the input file input.yml:

system: 'poses*.pdb' # Protein-ligand poseS!
chain: 'L' # Ligand chain ID
resname: 'THC' # Ligand residue name
seed: 12345
# Distance between two atoms to track the simulation
atom_dist:
- "A:2:CA" # First atom (chain ID:residue number:atom name)
- "B:3:CG" # Second atom
cpus: 60
rescoring: true # only relaxation part of the algorithm

For more optional flags please refer to optional flags.

3. Run simulation

To run the system launch the simulation with the following command:

python -m pele_platform.main input.yml

4. Output

Raw output

Trajectory and report files for each simulation are located in working_folder/output. That’s where you can find detailed information on each snapshot (PDB file, binding energy, metrics, etc.).

Selected poses

Clusters

Upon completion of the simulation, all trajectories are clustered based on ligand heavy atom coordinates. Then, a cluster representative with the best binding energy (or metric of your choice) is selected. Ranked cluster representatives can be found in:

working_folder/results/clusters

Best snapshots

In addition, top 100 structures with the best binding energy (or metric of your choice) are retrieved. This is done to ensure the clustering algorithm did not skip any valuable results. They are stored in:

working_folder/results/top_poses