[WIP] Plan rbfe #95
[WIP] Plan rbfe #95
Conversation
|
| if leg == 'complex': | ||
| rbfe_protocol = RelativeHybridTopologyProtocol(settings=complex_rfe_settings) | ||
| else: | ||
| rbfe_protocol = RelativeHybridTopologyProtocol(settings=ligand_rfe_settings) | ||
|
|
||
| transformation = openfe.Transformation( | ||
| stateA=system_a, | ||
| stateB=system_b, | ||
| mapping=new_edge, | ||
| protocol=rbfe_protocol, | ||
| name=name | ||
| ) | ||
| transformations.append(transformation) | ||
|
|
||
| return transformations |
There was a problem hiding this comment.
See here for how you can use the adaptive settings class on the RBFE protocol to do this for you, I would just copy what it does extract the settings from the protocol adapt them and then make a new protocol and Transformation. You might need to use the isinstance check as well if all RBFE protocols should go through this function.
| system. | ||
| name : str | ||
| The name / identifier of the benchmark system | ||
| network : LigandNetwork |
There was a problem hiding this comment.
Do we want to make it clear that this is for RBFE only, as absolute calculations wont need this network file?
| settings.simulation_settings.time_per_iteration = 2.5 * unit.picosecond | ||
| settings.simulation_settings.real_time_analysis_interval = 1 * unit.nanosecond | ||
| settings.output_settings.checkpoint_interval = 1 * unit.nanosecond | ||
| settings.solvation_settings.box_shape = 'dodecahedron' | ||
| settings.forcefield_settings.nonbonded_cutoff = 0.9 * unit.nanometer | ||
| settings.solvation_settings.solvent_padding = 1.0 * unit.nanometer # for complex | ||
| #settings.solvation_settings.solvent_padding = 1.5 * unit.nanometer # for ligand |
There was a problem hiding this comment.
No longer needed this will be done automatically by adaptive settings.
| #settings.solvation_settings.solvent_padding = 1.5 * unit.nanometer # for ligand | ||
|
|
||
| # Only run one repeat per input json file | ||
| settings.protocol_repeats = 1 |
There was a problem hiding this comment.
Good idea lets keep this!
| """ | ||
|
|
||
| @staticmethod | ||
| def _protocol(forcefield) -> RelativeHybridTopologyProtocol: |
There was a problem hiding this comment.
Do you imagine there being different scripts for the different protocol types, which act in a similar way for example, hybridtopology, non-equilibrium cycling/switching and SepTop or could we just expose a flag to chose one of these types of RBFE protocol and reuse a lot of the script?
There was a problem hiding this comment.
I envisioned that the BenchmarkSystem class has most of the meat, but the RBFEBenchmarkSystem will be different in overwriting the _process_inputs and the proper _protocol methods. I'll have to see how the adaptive settings would interact with the _protocol method, but it's abstract in the base class so that it can be specific for the calculation type.
|
|
||
| class BenchmarkSystem(ABC): | ||
| """ | ||
| Abstract base class defining the components of a and alchemical network. |
There was a problem hiding this comment.
Something seems to be missing here.
| self.ligand_dict = ofebu.process_sdf(ligands[self.partial_charge_scheme], return_dict=True) | ||
|
|
||
|
|
||
| def _process_protein(self, protein): |
There was a problem hiding this comment.
We're currently working on adding membrane support to openfe protocols. That one would require loading the PDB into a ProteinMembraneComponent. Just as an FYI, we can deal with that when we get there =)
|
|
||
| @staticmethod | ||
| def _protocol(forcefield) -> RelativeHybridTopologyProtocol: | ||
| """Utility method for getting RFEProtocol settings for non charge changing transformations. |
There was a problem hiding this comment.
Since this returns settings, should this method be called _settings?
There was a problem hiding this comment.
sure! the class types are called protocols, but I see they are accessed with "settings"
| # if abs(charge_difference) > 1e-3: | ||
| # # Error out - this shouldn't be happening | ||
| # raise RuntimeError("charge change found!") | ||
| complex_rfe_settings = protocol |
There was a problem hiding this comment.
See comment above, these are settings and not a protocol, correct?
There was a problem hiding this comment.
Is there a difference? The settings appear to be contained in Protocol classes.
Close #89
Close #53
BenchmarkSystemand derived classRBFEBenchmarkSystemopenfe_benchmarks.dataPlease feel free to be nitpicky about naming and terminology.