Bristol University Docking Engine Force Field (BUDEFF)

BUDEFF is a standalone implementation of the BUDE (Bristol University Docking Engine) all-atom force field^1,2. The force field is developed by the Sessions group.

Installation

You can install BUDEFF from pip:

pip install budeff

Or from source by downloading/cloning this repository, navigating to the folder and typing:

pip install .

BUDEFF uses Cython, so if you're installing from source make sure you have it installed.

Usage

The BUDE force field can be used to calculate energies for any protein structure that has been loaded into AMPAL, a simple framework for representing biomolecular structure. You can load a structure into AMPAL like so:

import ampal
structure = ampal.load_pdb('3qy1.pdb')

Once the structure is loaded in, you can now run BUFF on the structure. BUFF has two modes:

1. Internal Energy - In this mode a single AMPAL object is supplied and the energy between every pair of atoms is calculated, so long as the atom is parameterised in the force field.
2. Interaction Energy - In this mode a list of AMPAL objects is supplied and the energy of atom pairs forming the interaction between these objects is measured. For example, if the interaction energy between object A and object B, then all atom pairs will contain one atom from A and one from B.

import budeff
internal_energy = budeff.get_internal_energy(structure)
# OUT: NotParameterisedWarning: O (HOH) atom is not parameterised in the selected residue force field.
# OUT:   warnings.warn(w_str, NotParameterisedWarning)

While the score was being calculated, a NotParameterisedWarning was raised. This tells us that the water (HOH) is not parameterised in the force field and so will be ignored. The BUDE force field has been developed for performing protein docking, and so only protein and a few common ions are parameterised.

get_internal_energy returns a BuffScore object:

print(internal_energy)
# OUT: <BUFF Score -7108.00: 214.37 St | -4343.46 De | -2978.91 Ch>

The BuffScore contains information on the total energy of the system (-7108.00 in this case) as well as the different components of this score, which are steric (214.37 St), energy of desolvation (-4343.46 De) and charged interactions (-2978.91 Ch). Each of these components can be accessed individually:

print(internal_energy.total_energy, internal_energy.steric,
      internal_energy.desolvation, internal_energy.charge)
# OUT: -7108.000086377617 214.36602045772776 -4343.460484501997 -2978.905622333365

Individual pairwise interactions can be examined. The inter_scores attribute is a list of all the pairwise interactions with non-zero scores that are used to create the score:

print(internal_energy.inter_scores[0])
# OUT: ((<Carbon Atom (CA). Coordinates: (15.518, -30.153, -25.207)>,
# OUT:   <Carbon Atom (CB). Coordinates: (17.842, -27.509, -21.862)>),
# OUT:  [0.0, -0.10352520993045879, 0.0])

Each element in inter_scores contains a pair of atoms which form the interaction and a list with the different elements of the scoring function in the order steric, desolvation and charge.

To calculate the interaction energy, use the get_interaction_energy function. This take a list of ampal objects and calculates the interaction energy between these objects:

interaction_energy = budeff.get_interaction_energy([structure[0], structure[1]])
print(interaction_energy)
# OUT: NotParameterisedWarning: O (HOH) atom is not parameterised in the selected residue force field.
# OUT:   warnings.warn(w_str, NotParameterisedWarning)
# OUT: <BUFF Score -479.44: 26.19 St | -416.31 De | -89.32 Ch>

The score is lower in this case as only the energy between an atom in chain a and an atom in chain b is considered.

There's lots more functionality in the BUFF module so have a dig around.

References

1. McIntosh-Smith S. et al. (2012) Benchmarking energy efficiency, power costs and carbon emissions on heterogeneous systems. Comput. J., 55, 192–205.
2. McIntosh-Smith S. et al. (2014) High performance in silico virtual drug screening on many-core processors. Int. J. High Perform. Comput. Appl., 29, 119–134.