meld.runner.transform.TransformerBase

class meld.runner.transform.TransformerBase(param_manager, mapper, density_manager, builder_info, options, always_active_restraints, selectively_active_restraints)

Bases: object

Base class to document how transformers work.

The transformation process proceeds in several stages.

• The transformer is initialized with the run options and lists of restraints. The transformer must store any relevant information, as these parameters will not be given again. The __init__ method must also delete any restraints that are handled by this transformer from always_active_restraints and selectively_active_restraints.

• Next, add_interactions is called. This provides an opportunity for the transformer to add new forces, or to modify or replace existing forces.

• Next finalize is called. All changes to particles and forces are completed before this call, which can be useful when we have to, e.g. store parameters.

• Finally, update is called every stage. This provides an opportunity to update parameters like force constants depending on alpha or on the time step.

The number of particles cannot be changed by a transformer, instead additional particles should be added using a patcher.

__init__(param_manager, mapper, density_manager, builder_info, options, always_active_restraints, selectively_active_restraints)

Initialize a Transformer

Parameters

• param_manager (ParameterManager) – parameter manager to handle sampling of paramters

• builder_info (dict) – information from the system builder / patcher

• options (RunOptions) – the options for the runner

• always_active_restraints (List[Restraint]) – these restraints are always active

• selectively_active_collections – these restraints are selected by the MELD algorithm

Methods

__init__(param_manager, mapper, ...)	Initialize a Transformer
add_interactions(state, system, topology)	Add new interactions to the system.
finalize(state, system, topology)	Finalize the transformer.
update(state, simulation, alpha, timestep)	Update the system according to alpha and timestep.

add_interactions(state, system, topology)

Add new interactions to the system.

This may involve:

• Adding new forces, e.g. for restraints

• Replacing an existing force with another, e.g. softcore interactions

This method must return the modified system. If the transformer does not add interactions, it may simply return the passed values.

Parameters

• state (IState) – the state of the system

• system (System) – OpenMM system object to be modified

• topology (Topology) – OpenMM topology object to be modified and/or used for indexing

Return type

System

finalize(state, system, topology)

Finalize the transformer.

This method is guaranteed to be called after all forces are added to the system and provides an opportunity to do bookkeeping.

This method should not add any new forces.

Parameters

• state (IState) – the state of the system

• system (System) – OpenMM system object to be modified

• topology (Topology) – OpenMM topology object to be modified and/or used for indexing

Return type

None

update(state, simulation, alpha, timestep)

Update the system according to alpha and timestep.

This method is called at the beginning of every stage. It should update forces and parameters as necessary.

Parameters

• state (IState) – the state of the system

• simulation (Simulation) – OpenMM simulation object to be modified

• alpha (float) – current value of alpha, ranges from 0 to 1

• stage – current stage of the simulation, starting from 0

Return type

None