Equilibration and Non Equilibrium Switching [MDPrep] [MDRun] [FECalc]

Category Paths

  • Product-based/Molecular Dynamics/GROMACS

  • Product-based/Molecular Dynamics/OpenMM

  • Role-based/Computational Chemist

  • Role-based/Medicinal Chemist

  • Task-based/Molecular Dynamics

  • Task-based/Affinity Prediction

  • Solution-based/Hit to Lead/Affinity Prediction/Free-Energy Calculations

  • Solution-based/Small Molecule Lead-opt/Affinity

Description

  • Purpose:

    • This Floe performs relative binding free energy (RBFE) calculations using the nonequilibrium switching (NES) method refined by the de Groot lab (Gapsys et al., Chem. Sci., 2020, 11, 1140-1152). It also carries out the equilibration MD runs which must precede NES.

  • Method Recommendations/Requirements:

    • Four inputs are required:

      • A protein prepared to MD standards: protein chains must be capped, all atoms in protein residues (including hydrogens) must be present, and missing protein loops resolved or capped. Crystallographic internal waters should be retained where possible.

      • A dataset of posed ligands need to have reasonable 3D coordinates, all atoms, and correct chemistry (in particular bond orders and formal charges). The starting poses should not have very high gradients, in particular no bad clashes with the protein.

      • A text file of edges (explained below), one line per edge, of form “ligA_name >> ligB_name”.

      • [Optional] a text file containing experimental binding free energies for at least one ligand, one experimental datapoint per line, of form “ligA_name {deltaG(exptl)} {error_deltaG(exptl)} {units}” for example, “gn1c -8.56 0.17 kcal/mol”.

  • Limitations

    • If no experimental binding free energies (the fourth input above) are given, the estimation of ligand binding free energies has no reference value so the relative values will be centered around the mean.

    • Currently there is no mitigation for the effects of changes in buried waters, protein sidechain flips, or large protein movements between ligA and ligB.

  • Expertise Level:

    • Regular

  • Compute Resource:

    • High

  • Keywords:

    • MDPrep, MD, FECalc

  • Related Floes:

    • Ligand Bound and Unbound Equilibration for NES [MDPrep] [MD]

    • Non Equilibrium Switching [MD] [FECalc]

    • Compare Experimental Affinity with NES Results [Utility] [FECalc]

    • Non-Equilibrium Switching Recovery [Utility] [FECalc]

This Floe combines Equilibration MD calculations of the bound and unbound ligands with subsequent Relative Binding Free Energy calculations using Non-Equilibrium Switching. Given the inputs of the protein and posed ligands, the complex is formed with each ligand/conformer separately, and the bound and unbound simulations are then carried out. Each ligand can have multiple conformers but each conformer will be run separately as a different ligand. Currently only one of the conformers will be used in the NES calculations. A minimization stage is performed on the system followed by a warm up (NVT ensemble) and several equilibration stages (NPT ensemble). In the minimization, warm up, and equilibration stages, positional harmonic restraints are applied on the ligand and protein. At the end of the equilibration stages a production run (by default 6 ns) is performed on the unrestrained system. Separate datasets are written for the bound and unbound ligands.

Then, Relative Binding Free Energy (RBFE) calculations are performed using the Non-Equilibrium Switching (NES) method. Here the third input mentioned above is used, a the text file of edges, describing the map of desired alchemical transformations of one ligand into another. Each transformation forms an edges of a connected graph of ligands. The file must have one line per transformation, of format

ligA_name >> ligB_name

where “ligA_name” and “ligB_name” are the ligand names for the ligands to be transformed. These ligand names must correspond exactly to those in the bound and unbound ligand equilibration datasets.

The Floe will draw a number of starting snapshots from the bound and unbound trajectories of the ligands. Then for each edge in the edge file, it will generate an RBFE alchemical transformation from ligA into ligB, and carry out the NES fast transformation of ligA into ligB, and vice versa, for each of the snapshots. The resulting relative free energy change, or DeltaDeltaG, for each edge is the primary output of this method. A maximum likelihood estimator is then used to derive a predicted binding affinity (free energy, or DeltaG) for each ligand. The mean value of the input experimental binding free energies is used as the reference value for the computed ones.

The speed of the NES transformation and the number of snapshots transformed can be adjusted from default values by the user at runtime. The floe outputs two floe report/dataset pairs, one for the calculated RBFE edges (DeltaDeltaGs), and one for the derived affinity predictions (DeltaGs) of ligand.

Promoted Parameters

Title in user interface (promoted name)

Inputs

Protein Input Dataset (protein): Protein Input Dataset

  • Type: data_source

Ligand Input Dataset (ligands): Ligands-only input dataset or protein-ligand input dataset containing Design Unit prepared by SPRUCE

  • Required

  • Type: data_source

Mapper Dataset (mapper): Mapper Input Dataset

  • Required

  • Type: data_source

CPU GPU Spot Policy Selection

CPUs (cpu_count_md): The number of CPUs to run this cube with

  • Type: integer

  • Default: 16

GPUs (gpu_count_md): The number of GPUs to run this cube with

  • Type: integer

  • Default: 1

Spot policy (spot_policy_md): Control cube placement on spot market instances

  • Type: string

  • Default: Required

  • Choices: [‘Allowed’, ‘Preferred’, ‘NotPreferred’, ‘Prohibited’, ‘Required’]