More info/docs on coloring or selections?

[davidlmobley]

BTW, this is great. Beautiful smooth viewer.

But, now that I have this working on my mixed solvent system (as discussed here: https://github.com/mdtraj/mdtraj/issues/1122#issuecomment-222601126) I run into the issue that I'd like to be able to control the viewer to make, say, phenol molecules be one color and cyclohexane be another.

It is unclear to me how to do this from the info I can find easily. I find code like this in the README.md:

view.add_cartoon("protein", color="residueindex")
view.add_surface("protein", opacity=0.3)

But I have no protein, and it's unclear to me what is acceptable for the first argument. If I try help(view.add_cartoon) it only gives me: func(this, selection='all', **kwargs) method of nglview.NGLWidget instance

which is unhelpful. And, trying view.add_cartoon("phenol", color="residueindex") makes my viewer window disappear without any error message or anything (bug?).

[arose]

what are the group names of your phenol/cyclohexane molecules?

[hainm]

Can you try to replace 'phenol' by its resname?

view.add_surface('ASP', opacity=0.3)

Please see selection language here

http://arose.github.io/ngl/doc/#User_manual/Usage/Selection_language

We are working in writing more doc now, but feel free to ask anything here.

[hainm]

Change color

view.add_surface('1', color='blue')

[davidlmobley]

My phenol has residue name ZTF, it seems (these are actually assigned randomly in this case). However, coloring it seems to fail, i.e. view.add_surface('ZTF', color='blue') results in no viewer window appearing (with no error) and likewise view.add_surface('1', color='blue')

I'll consult the selection language and see if I can do anything useful.

[davidlmobley]

Hmm, so it seems not to be a problem with my selections, in that I don't get any viewer window even if I just do: view.add_surface('all', opacity=0.3) which seems like it ought to work.

(Also, again, I get no info at all - just no viewer appearing, and no error messages or anything).

[davidlmobley]

I have the same issue (viewer window disappears) if I try: view.add_representation('spacefill', selection='all' ) or any variant I've been able to come up with.

[davidlmobley]

@hainm - I also tried the same experiment in the test_mdtraj_nglview environment we created in the other thread, but I have the same issue - any messing with representation/surface/etc at all causes the viewer window to disappear.

I also have two related problems with my other notebook which tries to view a simple trajectory run with OpenMM:

• If I view a trajectory run with OpenMM, I get an empty viewer window with a "play" button at the bottom but there is no content to play
• In troubleshooting, if I just remove that trajectory and instead view via view = nglview.show_pdbid("3pqr"), it works, UNTIL I try to change the representation at all, such as via view.add_representation('spacefill', selection='all' ) in which case the viewer window just doesn't come up (no error message, etc).

I'm guessing this means I should probably find something else to show in my class (which was supposed to be today but I switched the lecture to Thursday so I'd have time to deal with this) as this seems maybe not quite ready for general usage, unless I'm missing something obvious.

[hainm]

Can you attach a simple notebook so i can look?

[davidlmobley]

The one(s) in teh other thread should work but I'll re-attach.

[hainm]

Btw, in my example 'ASP' is Aspartic residue. :d what I really meant is to replace it with whatever name that mdtraj/openmm give you.

[davidlmobley]

(About "ASP", I mistyped my code. I'd tried to select "ZTF" but it failed).

[hainm]

Ok, i will check your nb in about 1-2 hours later.

[davidlmobley]

debugging.zip

OK, please try this. Maybe try Lec17_mixtures.ipynb first. Note the box at the top should be run to import stuff, but will fail with an exception because of the existing files I've included in 'data'. AMBER prmtop and .crd files are there in the tleap directory. I've also tried loading GROMACS ones but I can't get that to work for some reason.

[davidlmobley]

So, er, to get it to work, try making the top two lines:

traj = md.load( 'data/tleap/phenol_cyclohexane_3_100.inpcrd', top = 'data/tleap/phenol_cyclohexane_3_100.prmtop' )
#traj = md.load( 'data/gromacs/phenol_cyclohexane_3_100.gro', top = 'data/gromacs/phenol_cyclohexane_3_100.gro' )

and then removing the line view.add_representation('1', opacity=0.3)

If you add that last line back in, the viewer will disappear again (at least if it works for you the way it does for me).

[hainm]

@davidlmobley The first thing I realize is that you never display view to make it work

In first cell

view = nglview.show_mdtraj(traj)
#view = nglview.show_pytraj(traj_pt)
view # you need to call this to display

then in 2nd cell

view.add_surface(...)
view.add_licorice(...)

I will look other issues.

[davidlmobley]

The code in my notebook has:

view = nglview.show_mdtraj(traj)
#view = nglview.show_pytraj(traj_pt)
view

[davidlmobley]

Elaborating, the relevant box has (stripping out commented stuff):

import mdtraj as md
traj = md.load( 'data/tleap/phenol_cyclohexane_3_100.inpcrd', top = 'data/tleap/phenol_cyclohexane_3_100.prmtop' )

import nglview
import pytraj as pt

view = nglview.show_mdtraj(traj)
view

#view.add_representation('1', opacity=0.3)

If I leave it as is, I get a viewer with useful info in it. If I comment out the last line, I get nothing (no view window, no error, nothing).

[hainm]

yes, that is my point. To display in notebook, it need to call view in the last command.

[hainm]

You need to add view.add_representation('1', opacity=0.3) in another notebook cell.

[hainm]

[davidlmobley]

Ah, so you're saying I just have to move the change of representation up.

Well, that wasn't clear at all from the docs or examples. But yes, that seems to work for this.

OK, so that solves the problem with my mixed solvent notebook. However, I still have the issue that my OpenMM simulation brings up an empty viewer.

[davidlmobley]

(Thanks for that example you just sent though! Extremely helpful. So nwo I can use this for mixed solvents. Now if only I could view trajectories...)

[hainm]

OK, so that solves the problem with my mixed solvent notebook. However, I still have the issue that my OpenMM simulation brings up an empty viewer.

can you make a list of bullet points in your 1st post so I can easily check?

[hainm]

this is your residue names

print("residue set: %s" % set(residue.name for residue in traj.topology.residues))
# residue set: {'pheno', 'cyclo', 'solut'}

so, you can use

view.add_surface('pheno', opacity=0.3)

[davidlmobley]

I'm going to have to disappear for a couple of hours momentarily. Bullet points on what?

Here's the code

import os
import mdtraj
import mdtraj.reporters
from simtk import unit
import simtk.openmm as mm
from simtk.openmm import app
import mdtraj.testing
#pdb = mdtraj.load(mdtraj.testing.get_fn('native.pdb'))
pdb = mdtraj.load_pdb('http://www.rcsb.org/pdb/files/1LE1.pdb')
topology = pdb.topology.to_openmm()
forcefield = app.ForceField('amber99sbildn.xml', 'amber99_obc.xml')
system = forcefield.createSystem(topology, nonbondedMethod=app.CutoffNonPeriodic)
integrator = mm.LangevinIntegrator(330*unit.kelvin, 1.0/unit.picoseconds, 2.0*unit.femtoseconds)
simulation = app.Simulation(topology, system, integrator)
simulation.context.setPositions(pdb.xyz[0])
simulation.context.setVelocitiesToTemperature(330*unit.kelvin)
if not os.path.exists('test.h5'):
    simulation.reporters.append(mdtraj.reporters.HDF5Reporter('test.h5', 10))
    simulation.step(100)
print("Done")

from __future__ import print_function
import nglview
from IPython import display

traj = mdtraj.load('test.h5')
print(traj)

view = nglview.show_mdtraj(traj)
view

and it yields just an empty view window.

I'm not totally sure I'm doing the OpenMM simulation right, but this DID work a year ago with TrajectoryView, so one would think it should work. Really, I'm just after a minimal example of visualizing ANY OpenMM simulation, so if you don't like this one I'm open to anything else.

[hainm]

Bullet points on what?

never mind. (I just want you make like this)

• [ ] phenol selection ...
• [ ] openmm issue ...

- [ ] phenol selection ...
- [ ] openmm issue ...

[hainm]

Per openmm issue, it's likely due to openmm and mdtraj themself. If I ran you script in my notebook with py35. I got

/Users/haichit/miniconda/lib/python3.5/site-packages/tables/file.py in open_file(filename, mode, title, root_uep, filters, **kwargs)
    302                 raise ValueError(
    303                     "The file '%s' is already opened, but "
--> 304                     "not in read-only mode (as requested)." % filename)
    305             # 'a' and 'r+' are compatible with everything except 'r'
    306             elif mode in ('a', 'r+') and omode == 'r':

ValueError: The file 'test.h5' is already opened, but not in read-only mode (as requested).

[hainm]

After running openmm in one notebook, then open another one to reload test.h5 file, it's ok too me.

[hainm]

more fun stuff

[davidlmobley]

In Python 2.7, it doesn't work for me even if I just try to load the trajectory; after the usual import stuff at the top (cut off) I get this:

[davidlmobley]

Maybe send me your traj.h5 and I can check if it's a problem with that?

[davidlmobley]

I could also go back to trying the minimal test environment you created earlier.

I had a couple minutes before I run out the door, hence the responses. But I'll (shortly) be delayed by a while.

[hainm]

sorry, just had dinner. This test.h5 works for me. test.h5.zip

[hainm]

also try files from nglview

this is also a trajectory

[davidlmobley]

Well, yup, your test.h5 works for me. So something is wrong with mine.

That narrows things down by about, well, 50%. :) Any advice? Otherwise I'll go work on my openmm knowledge and bug those guys. Thanks for your help!

[hainm]

Otherwise I'll go work on my openmm knowledge and bug those guys. Thanks for your help!

this definitely should go to openmm issue, out of our league since we (me and @arose) are not using openmm.

[hainm]

btw, I just install openmm and mdtraj from omnia

conda install openmm -c omnia --force
conda install mdtraj -c omnia --force

[hainm]

I just update a bit about coloring and atom selection: https://github.com/arose/nglview/commit/d90f57a3575e222babfacfa443ebd0923300c402

Due to our limited time for this (nglview) side project, doc is never done thoroughly. But please ask anything you need, we might have it in nglview but hidden ( :D ).

For example, you can even add more mdtraj.Trajectory as you like

# 1st cell
view = nglview.show_mdtraj(traj0)
view

# 2nd cell
# wrap mdtraj.Trajectory in nglview object
view.add_trajectory(nglview.MDTrajTrajectory(traj2))

# you can even mix pytraj and mdtraj too
view.add_trajectory(nglview.PyTrajTrajectory(pytraj_object))

# or even rdkit :D 
view.add_trajectory(nglview.RdkitStructure(its_object))

Of course, you need to take care by your own trajectory superimpose in mdtraj.

[davidlmobley]

Thanks for all the help! Turns out I have my issues resolved now. It appears the main problem was that I was just using an old .h5 file that wasn't being overwritten when I ran new ones because of the if statement. :) So now everything works. Nice viewer - good work! I'll definitely use this again in the future.

[hainm]

cool.

[davidlmobley]

So, @hainm - if I DO try:

view.add_surface('pheno', opacity=0.3)

for some reason I get a surface which seems to correspond to the surface of my box rather than the surface of the individual phenol molecule. Insights?

[hainm]

can you post screenshot?

[davidlmobley]

Also, relatedly, if I clear all representations and then do view.add_licorice('solut', color="blue") or view.add_licorice('pheno', color="blue")

I end up with everything colored blue...

This is with my residue set as residue set: set(['pheno', 'solut', 'cyclo']).

I'd really like to be able to color things separately and visualize just certain tings.

[davidlmobley]

The surface issue. Zoomed in:

Zoomed out:

Sorry, I'm sure all of this is just me doing something wrong. But, maybe it will help you in terms of knowing the minimum you have to explain to a novice to get them going (i.e. the "view command has to be last" was a key piece of info which was non-obvious).

[hainm]

you need to set opacity=0.3 (or others). The default is 1.0.

[hainm]

(i.e. the "view command has to be last" was a key piece of info which was non-obvious).

this is my English issue.

[davidlmobley]

No, no. The problem is that I should be selecting only one molecule! i.e., "solut" specifies a single residue/moelcule.

[davidlmobley]

In other words, what I should be seeing is a single molecule represented as a surface and everything else should be licorice.

[hainm]

hold on. Let me check myself.

[davidlmobley]

Ah, OK. It seems as though it's the selection by residue name which is not working. If I use residue numbers I get the expected result: