Add reproducible version of COVID Moonshot example anyone can run as an example #1145
Conversation
| # Modeller needs topology and positions. Lots of trial and error found that this is what works to get these from | ||
| # an openforcefield Molecule object that was created from a RDKit molecule. | ||
| # The topology part is described in the openforcefield API but the positions part grabs the first (and only) | ||
| # conformer and passes it to Modeller. It works. Don't ask why! |
There was a problem hiding this comment.
is there any documentation we can refer to here?
| simulation.reporters.append(XTCReporter(arguments['--xtctraj'], reporting_interval, atomSubset=output_indices)) | ||
|
|
||
| # Add reporter to generate PDB trajectory, if requested | ||
| # NOTE: The PDBReporter does not currently support atom subsets |
There was a problem hiding this comment.
is there any way to do this in the future? For ligand<->residue contacts just writing out the ligand + selected residues would be super useful..
There was a problem hiding this comment.
Yes, it should be easy--I've opened an issue: openmm/openmm#3925
| output_positions = context.getState(getPositions=True, enforcePeriodicBox=False).getPositions(asNumpy=True) | ||
| with open(arguments['--final'], 'w') as outfile: | ||
| PDBFile.writeFile(output_topology, output_positions[output_indices,:], file=outfile, keepIds=False); | ||
|
|
There was a problem hiding this comment.
would it be in scope here to implement writing out some 'representative' snapshots, i.e. do a quick clustering of the trajectory and write cluster centers to PDBs? Something like https://mdtraj.org/1.9.3/examples/centroids.html
But then again, not sure how well that scales to PL systems.
There was a problem hiding this comment.
(asking because this is an ASAP request -> would like to see quantitatively how much ligand sections like to reside in specific subpockets)
There was a problem hiding this comment.
Potentially, but I'm not certain yet whether we need to keep this particular script here.
|
|
||
| """ | ||
|
|
||
| cdd_csv_filename = 'CDD CSV Export.csv' |
There was a problem hiding this comment.
| cdd_csv_filename = 'CDD CSV Export.csv' | |
| cdd_csv_filename = 'CDD CSV Export.csv' | |
| TODO: implement path into CLI |
| # Drop NaNs | ||
| print(f'Dropping NaNs...') | ||
| df.dropna(axis=0, how='any', thresh=None, subset=None, inplace=True) | ||
| print(f'{len(df)} records remain') |
There was a problem hiding this comment.
we should add an exception here in case len(df) == 0
There was a problem hiding this comment.
@JenkeScheen : Could you help generalize this and migrate it to https://github.com/choderalab/covid-moonshot-ml?
There was a problem hiding this comment.
IIRC @kaminow is doing this for the moonshot manuscript?
There was a problem hiding this comment.
@jchodera @JenkeScheen I've just opened a new PR for this: asapdiscovery/asapdiscovery#130
| pred = oechem.OEAtomMatchResidue(["GLN:189: :A"]) | ||
| protonate_opts = opts.GetPrepOptions().GetProtonateOptions(); | ||
| place_hydrogens_opts = protonate_opts.GetPlaceHydrogensOptions() | ||
| place_hydrogens_opts.SetNoFlipPredicate(pred) |
There was a problem hiding this comment.
This is giving me downstream issues with OpenMM complaining about protonation states. For some PDBs (I can provide exmaples if needed) this set of place_hydrogens_opts produces wonky forms of histidine (e.g. all single bonds for some reason?). Using this block instead:
pred = oechem.OEAtomMatchResidue(["GLN:189:.*:.*:.*"])
protonate_opts = opts.GetPrepOptions().GetProtonateOptions();
place_hydrogens_opts = protonate_opts.GetPlaceHydrogensOptions()
place_hydrogens_opts.SetBypassPredicate(pred)
place_hydrogens_opts.SetNoFlipPredicate(pred)
protonate_opts = oespruce.OEProtonateDesignUnitOptions(place_hydrogens_opts) # ?
opts.GetPrepOptions().SetProtonateOptions(protonate_opts) # ?
Fixes most of the issues. Perhaps @kaminow or @apayne97 can comment on the best practice here?
|
Merging this example because it's needed for the manuscript. |
jchodera
changed the title
This PR adds examples from the COVID Moonshot that anyone should be able to run.
This is currently a work in progress, but includes:
moonshot-mainseries/- relative binding free energy calculations primary synthetic series milestones COVID Moonshotmoonshot-sulfonamides/- relative binding free energy calculations of a congeneric sulfonamide series from the COVID Moonshot from a single structure