[TUTORIAL] Funnel metadynamics

[lohedges]

This is a thread to discuss the creation of a tutorial showing how to implement funnel metadynamics within BioSimSpace.

[jmichel80]

Adding current diagram summarising the use case:

[dlukauskis]

I've ran a fun-metaD sim that was setup for gromacs and I got a few errors.

1. Missing definition for the ligand center of mass, after the definition of p1 and p2 points.
   lig: COM ATOMS=first_ligand_atom_id-last_ligand_atom_id

2. Missing sigma parameter for the extent CV, so instead of
   metad: METAD ARG=pp.proj,pp.ext SIGMA=0.025 ...

   it should be

   metad: METAD ARG=pp.proj,pp.ext SIGMA=0.025,0.05 ...

[lohedges]

Thanks for catching these. I'll be able to push a fix tomorrow.

…

On Mon, 22 Mar 2021, 13:02 dlukauskis, ***@***.***> wrote: I've ran a fun-metaD sim that was setup for gromacs and I got a few errors. 1. Missing definition for the ligand center of mass, after the definition of p1 and p2 points. lig: COM ATOMS=first_ligand_atom_id-last_ligand_atom_id 2. Missing sigma parameter for the extent CV, so instead of metad: METAD ARG=pp.proj,pp.ext SIGMA=0.025 ... it should be metad: METAD ARG=pp.proj,pp.ext SIGMA=0.025,0.05 ... — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub <#194 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAE6K3IIUF5ATUZHWHMVRDDTE45W3ANCNFSM4YPIREAQ> .

[dlukauskis]

I've made some functions that draw the funnel for NGLview, here's the code. The idea behind using NGLview is to rapidly check that the funnel points out the right way and includes/excludes the right bits of the protein. @lohedges, could you have a go at making my functions built into BSS code?

[lohedges]

Thanks, @dlukauskis. I've fixed the issues with the GROMACS PLUMED file and have added a check so that the ProjectionOnAxis.cpp file is only copied to the working directory when the PLUMED version is less than 2.7.

The plotting functionality looks great. I'll start implementing this now. Just to check... you define three vectors: origin, v1, and v2. Both origin and v1 are identical in this case. Which system did you use to define the values of v1 and v2? Are these just the COM of p0 and p1 used to define the funnel CV? (From reading the function doc strings, it would seem so.) I see you include a solvated.pdb file. Does this correspond to one of the systems from your original tutorial?

[dlukauskis]

Oh yeah, origin = v1 = p1 and v2 = p2. This was based on some older bit of code so I forgot to remove all of the redundancies like this. Solvated.pdb is 2WI3, which is almost identical to 2WI2 from the tutorial, except for a slightly different binding pose.

[lohedges]

I've now implemented the visualisation code natively in BioSimSpace. You can do the following to generate a BioSimSpace.Notebook.View object, e.g. using files from your tutorial:

import BioSimSpace as BSS

# Load the system.
system = BSS.IO.readMolecules(["input/2WI2.prmtop", "input/2WI2.rst7"])

# Create the funnel parameters.
p0, p1 = BSS.Metadynamics.CollectiveVariable.makeFunnel(system)

# Define the collective variable.
cv = BSS.Metadynamics.CollectiveVariable.Funnel(p0, p1)

# Create a view.
view = BSS.Metadynamics.CollectiveVariable.viewFunnel(system, cv)

This can then be used to visualise the funnel.

View the entire solvated system and funnel:

view.system()

View the protein, ligand, and funnel: (Assuming the protein and ligand are the first two molecules)

view.molecules([0, 1, -1])

Just view the funnel:

view.molecule(-1)

Let me know if I've missed anything, or if you come across any issues.

[dlukauskis]

Yep, that visualisation code works really well! I noticed one issue with the PLUMED file still, where line 473 in _plumed.py
colvar_string = "lig: COM=%d-%d" % (idx+1, idx+num_atoms)
should be
colvar_string = "lig: COM ATOMS=%d-%d" % (idx+1, idx+num_atoms)

I'm also attaching a metadynamics.py for OpenMM 7.4.2 that allows to print hill heights.
metadynamics.zip

[lohedges]

Thanks, I'll try to figure out how we can monkey-patch the version of OpenMM that we bundle (either in the binary or from conda). It looks like nothing has changed in the metadynamics code for OpenMM 7.5, so this should work for the conda-forge build of OpenMM too.