Interactive simulation #129

dane1122 · 2017-02-01T19:40:52Z

Pull request!

avirshup

Looks like a great start! Just a few more things need to get hooked up before we're ready to start running it.

avirshup · 2017-02-01T19:47:58Z

moldesign/integrators/lammps.py

+@exports
+class LAMMPSNvt(ConstantTemperatureBase):
+    def __init__(self, *args, **kwargs):
+        super(LAMMPSNvt, self).__init__(*args, **kwargs)


You can omit this __init__ method - the superclass __init__ will get called automatically if you do.

avirshup · 2017-02-01T19:50:22Z

moldesign/integrators/lammps.py

+        """
+        Users won't call this directly - instead, use mol.run
+        Propagate position, momentum by a single timestep using velocity verlet
+        :param run_for: number of timesteps OR amount of time to run for


Looks like this docstring came from a piece of code I haven't touched in a while :)

We've switched over to Google's docstring format - see http://google.github.io/styleguide/pyguide.html?showone=Comments#Comments

So you'll want to make sure to write your docstrings in that format (or update mine, as the case may be)

avirshup · 2017-02-01T19:52:34Z

moldesign/integrators/lammps.py

+        #     if istep + 1 >= next_trajectory_frame:
+        #         self.traj.new_frame()
+        #         next_trajectory_frame += self.params.frame_interval
+        # return self.traj


Once we're ready to get started with the run method, it will need to:

Send the molecule's current positions and momenta to LAMMPS

Create a Trajectory object to store the results.

Run dynamics in LAMMPS, storing new a frame to the Trajectory object at fixed simulation time intervals (defined in self.params.frame_interval )

avirshup · 2017-02-01T19:55:52Z

moldesign/integrators/lammps.py

+        lammps_system.command(nvt_command)
+
+        # TODO:
+        if self.params.constrain_hbonds && self.model.group_hbond == True :


No && or || in python - it's just and and or

avirshup · 2017-02-01T20:04:05Z

moldesign/models/lammps.py

+    def calculate(self, requests):
+
+        lammps_system = self.lammps_system
+        lammps_system.run(500)


This part should be in the integrator - we won't do any running in the EnergyModel

avirshup · 2017-02-01T20:18:39Z

moldesign/models/lammps.py

+        # Store the PRMTOP file if ff are assigned and use it to create lammps molecule data
+        self.mol.ff.amber_params.prmtop.put('/tmp/tmp.prmtop')
+
+        import parmed as med


This should go at the top of the function declaration

avirshup · 2017-02-01T20:19:27Z

moldesign/models/lammps.py

+        parmedmol = med.load_file('/tmp/tmp.prmtop')
+
+        lmp = lammps()
+        L = PyLammps(ptr=lmp)


Variable name should be lower case - something like pylmp would work here

avirshup · 2017-02-01T20:23:55Z

moldesign/models/lammps.py

+            self.group_water = False
+
+        self.lammps_system = L
+        print 'Created LAMMPS object'


We generally want to keep console logging like this pretty light, except in the following situations:

We made potentially unexpected changes to the Molecule object.

The function has a somewhat ambiguous effect that needs to be logged

We're in the middle of a long simulation and need to provide status updates

For OpenMMPotential, we log the system creation because it falls into case 2). OpenMM systems bind to a specific hardware interface - the CPU, an nVidia GPU, or a generic GPU - so we need to tell the user which resource the simulation will be using. But here, there's nothing ambiguous, so we don't need to log it.

avirshup · 2017-02-01T20:25:53Z

moldesign/models/lammps.py

+        print 'Created LAMMPS object'
+
+
+    def _group_hbonds(parmedmol):


Everyone is bad at this, me especially, but it's still good to write at least a one-line docstring, even for private functions, unless it's really really really obvious

avirshup · 2017-02-01T20:28:13Z

moldesign/models/lammps.py

+    # See http://parmed.github.io/ParmEd/html/index.html for ParmEd units
+
+    def _create_lammps_data(parmedmol):
+


You'll of course need actually open a lammps_data file-like object here. It would be great if we could pass this file directly through the LAMMPS API so that we don't need to write anything to disk, but that's not always possible unfortunately.

Fix visualization problems

…ems like it's broken

In an attemp to fix Trajectory, merging master

…nd due to small box size

Merging.

dane1122 · 2017-04-07T19:34:29Z

Hey Aaron, could we merge this into master so we can use it for interactive simulation MST app workflow? It would be great if you can review the code for this!

…ds to ensure that molecule can move around

avirshup

Hi Day - the LAMMPSPotential class in general looks good! A few Python style comments here (will tackle the integrator and the interactive model next)

avirshup · 2017-04-10T22:25:50Z

moldesign/models/amber2lammps.py

@@ -0,0 +1,945 @@
+#! /usr/bin/evn python2


Let's move this file to moldesign/external and import it from there, just for consistency

avirshup · 2017-04-10T22:28:47Z

moldesign/models/lammps_model.py

+import tempfile
+import os
+import sys
+from .amber2lammps import *


import * is discouraged in most contexts because it makes it hard to track down variables. This should be refactored to from ..external import amber2lammps, and all variables from that module should be referred to as amber2lammps.Lammps etc.

avirshup · 2017-04-10T22:31:17Z

moldesign/models/lammps_model.py

+        self.lammps_system = None # Basic LAMMPS system for this molecule
+        self.unit_system = None
+
+    def __exit__(self, exc_type, exc_value, traceback):


Have you checked to see if this method gets executed? I think you'll want to rename it to __del__ if you want it to run when this object is garbage-collected.

avirshup

A few things to look at for the LAMMPS integrator

avirshup · 2017-04-10T22:39:36Z

moldesign/integrators/lammps_integrator.py

+        lmps.command("timestep " + str(self.params.timestep.value_in(u.fs)))
+
+        # Set NVT settings
+        nvt_command = "fix {0} all nvt temp {1} {2} {3}".format(self.NAME_NVT_SIM,


Let's make sure to apply @ktbolt's fix for the Langevin integrator here (i.e., use nve but apply a Langevin thermostat)

avirshup · 2017-04-10T22:43:13Z

moldesign/integrators/lammps_integrator.py

+            lmps.command("unfix " + fix.get('name'))
+
+        # Create trajectory object
+        self.traj = Trajectory(self.mol, unit_system=self._model.unit_system)


Don't specify unit_system here - it's actually for output, not input, so MDT will set it automatically.

avirshup · 2017-04-10T22:53:07Z

moldesign/integrators/lammps_integrator.py

+        self._prepped = True
+
+    def _configure_system(self):
+        # Get lammps object from model


We'll need to make sure to tell LAMMPS to handle geometry constraints. Here are the 3 things that the interface will need to handle:

if self.params.get('constrain_hbonds', False) == True, then you'll need to set up SHAKE to constrain all bonds involving a hydrogen atom

if self.params.get('constrain_water', False) == True, then you'll need to set up SHAKE to constrain all bond lengths in any water molecules (residue name HOH or WAT)

You'll need to examine the list of constraints in self.mol.constraints. This is a list of objects that inherit from GeometryConstraint (see moldesign/geom/constraints.py), each of which describes a constraint that needs to be satisfied during MD.

Alternatively, you can just raise NotImplementedError in any/all of these cases :)

Fixed with ae97b6b

avirshup · 2017-04-10T22:59:58Z

moldesign-examples/Untitled1.ipynb

@@ -0,0 +1,1913 @@
+{


Please go ahead and move this notebook to moldesign/_notebooks, and then you can delete the rest of the moldesign-examples directory

moved in ae97b6b

avirshup · 2017-04-11T00:00:28Z

@dane1122 - a couple issues I'm running into:

The LAMMPS integrator throws an error when using the base LAMMPS potential model, e.g.:

import moldesign as mdt
from moldesign import units as u
mol = mdt.from_pdb('1yu8')
protein = mdt.assign_forcefield(mol)
protein.set_energy_model(mdt.models.lammps_model.LAMMPSPotential)
protein.set_integrator(mdt.integrators.LAMMPSNvt)
protein.run(5.0*u.ps)

Raises AttributeError: 'LAMMPSPotential' object has no attribute 'affected_atoms'

It looks like it's looking for an attribute that's in the Interactive potential class.

We'll need to disable LAMMPS' built-in logging when debugging == False so that the notebook doesn't get swamped with messages like the following:

Default LAMMPS logging (click to embiggen)

LAMMPS output is captured by PyLammps wrapper

Welcome to amber2lammps, a program to convert Amber files to Lammps format!

Reading /tmp/tmp.crd... default_name
12
0
done.
Reading /tmp/tmp.top... default_name
12
Reading Charges...
Reading Atomic Number...
Reading Atomic Masses...
Reading Atom Types...
Reading Excluded Atoms...
Reading Non-bonded Parameter Index...
Reading Residue Labels...
Reading Residues Starting Pointers...
Reading Bond Force Constants...
Reading Equilibrium Bond Values...
Reading Angle Force Constants...
Reading Equilibrium Angle Values...
Reading Dihedral Force Constants...
Reading Dihedral Periodicity...
Reading Dihedral Phase...
Reading 1-4 Electrostatic Scaling Factor...
Reading 1-4 Van der Waals Scaling Factor...
Reading Solty...
Reading LJ A Coefficient...
Reading LJ B Coefficient...
Reading Bonds which include hydrogen...
Reading Bonds which dont include hydrogen...
Reading Angles which include hydrogen...
Reading Angles which dont include hydrogen...
Reading Dihedrals which include hydrogen...
Reading Dihedrals which dont include hydrogen...
Reading Excluded Atom List...
Reading H-Bond A Coefficient, corresponding to r12 term for all possible types...
Reading H-Bond B Coefficient, corresponding to r10 term for all possible types...
Reading H-Bond Cut...
Reading Amber Atom Types for each atom...
Reading Tree Chain Classification...
Reading Join Array: Tree joining information
Reading IRotate...
(warning: File too large!) done.
Converting... (warning: Guessing at periodic box!) done.
Writing /var/folders/8d/k0qh4ksd5nv5b2vp6plfx6240000gp/T/tmpDdrFfg/data.lammps_mol... done.

avirshup · 2017-04-11T00:02:42Z

moldesign/models/lammps_model.py

+
+@exports
+class LAMMPSInteractive(EnergyModelBase):
+


For the purposes of getting the PR merged into MDT, we don't need to worry about interactive potentials (we'll deal with that in the web app). So let's leave this class out of the current PR, and merge it in later.

Should the "pair_style lj/charmm/coul/long 8.0 10.0 10.0" be changed to "pair_style lj/charmm/coul/charmm/implicit 8.0 10.0" or is this for something else?

I see. How do I leave it out in the current PR, do you know?

…zhi coords to ensure that molecule can move around" This reverts commit bc562c3. Revert "Lammps model and integrator" This reverts commit 4454121. Format using parmed instead of using amber2lammps.py since it is being maintained

dane1122 · 2017-04-11T00:54:47Z

I think I fixed the bugs that you listed. Could you please verify for me? All the fixes should be in either one of the commits:
ae97b6b
418f022

avirshup · 2017-04-11T15:55:44Z

Code is looking good! I'll play with it a bit when I get into the office. In the meantime, could you check on these two test failures?
test_lammps_shake_constrain_hbonds
test_minimization_reduces_energy

…esign/_notebooks directory

dane1122 · 2017-04-11T17:11:39Z

Update test: 0e4f1f4

Now directly access parmed object.

avirshup · 2017-04-18T21:21:44Z

@dane1122 ~~looks like we're getting a NameError at lammps_model.py:147~~

Oh right, this because LAMMPS isn't installed on Travis, and we're not building the docker container yet (future PR). So never mind on this one

dane1122 added 2 commits January 31, 2017 10:26

integrating LAMMPS into MDT

0a1e163

hello

d3c20d6

avirshup added the do not merge label Feb 1, 2017

avirshup self-requested a review February 1, 2017 19:41

avirshup reviewed Feb 1, 2017

View reviewed changes

dane1122 added 25 commits February 1, 2017 13:51

Created Dockerfile for LAMMPS

f47049a

added exmaples folder to test LAMMPS integration

a4beceb

Modified models file in accordance with the code review

7ef8548

example for lammps usage

c53274e

LAMMPS integration into model integrator

26c0bdf

Added docstring for describing functions inside LAMMPS integrator

b694311

Modified model and integrator

1941f51

Merge branch 'workflow_fixes' into interactive_simulation

80f74e4

Fix visualization problems

example

b680740

Remove it

4f7ad81

Merged with lammps branch to get yml file in this repository

b7d66bf

Modified LAMMPS model

902f015

Cleanup

0d1ebb6

Specified units

526d24f

Modified model and integrator. TODO: Trajectory for integrator. It se…

8507817

…ems like it's broken

Modified example files

6ade6dd

example

38fdc70

Merge branch 'master' into interactive_simulation

0345f16

In an attemp to fix Trajectory, merging master

Modified integrator and model

85fb4cf

Modified. TODO: Fix LAMMPS simulation error

aa2ddbe

Modified docker makefile for lammps

0487c04

Modified example

a5082a0

Modified xlo, xhi, ylo, yhi, zlo, zhi since the molecule wrapped arou…

0fc2b0d

…nd due to small box size

Modified shake

7e309f9

Merge remote-tracking branch 'origin/master' into interactive_simulation

55242fa

Merging.

dane1122 added 7 commits February 13, 2017 15:38

Interactive slide bar

06707ab

Trajectory way

2a7fe6f

Stretch factor for nearby atoms

a408c99

Removed print that was used for debugging

7566047

Modified model for better visual effects

c994e54

Fixed model

9b1118d

Modified for web assembly usage

0f5e422

avirshup mentioned this pull request Feb 21, 2017

Review LAMMPS integration PR Autodesk/molecular-simulation-tools#174

Open

dane1122 and others added 4 commits February 21, 2017 16:04

Removed export since not required anymore

23a22b5

fixed simulation boz

4e8813e

Create a function to possibly use for LAMMPS-related workflows

51cd6df

Lammps model and integrator

4454121

Set simulation box larger than atom xlo, xhi, ylo, yhi, zlo, zhi coor…

bc562c3

…ds to ensure that molecule can move around

avirshup reviewed Apr 10, 2017

View reviewed changes

avirshup reviewed Apr 11, 2017

View reviewed changes

Dayeong Lee added 2 commits April 10, 2017 17:26

Fixed bug: can still run dynamics with potential model

418f022

Dayeong Lee added 2 commits April 11, 2017 09:39

Removed moldesign-examples directory since examples should be in mold…

674d66f

…esign/_notebooks directory

Updated test

0e4f1f4

Dayeong Lee added 2 commits April 11, 2017 12:04

Merge branch 'master' into interactive_simulation

842c637

Fixed error caused by merging master.

7210046

Now directly access parmed object.

avirshup removed the do not merge label Apr 18, 2017

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Interactive simulation #129

Interactive simulation #129

dane1122 commented Feb 1, 2017

avirshup left a comment

avirshup Feb 1, 2017

avirshup Feb 1, 2017

avirshup Feb 1, 2017

avirshup Feb 1, 2017

avirshup Feb 1, 2017

avirshup Feb 1, 2017

avirshup Feb 1, 2017

avirshup Feb 1, 2017

avirshup Feb 1, 2017

avirshup Feb 1, 2017

dane1122 commented Apr 7, 2017

avirshup left a comment

avirshup Apr 10, 2017

avirshup Apr 10, 2017

avirshup Apr 10, 2017

avirshup left a comment

avirshup Apr 10, 2017

avirshup Apr 10, 2017

avirshup Apr 10, 2017

dane1122 Apr 11, 2017

avirshup Apr 10, 2017

dane1122 Apr 11, 2017

avirshup commented Apr 11, 2017 •

edited

avirshup Apr 11, 2017 •

edited

ktbolt Apr 11, 2017 •

edited

dane1122 Apr 11, 2017

dane1122 commented Apr 11, 2017

avirshup commented Apr 11, 2017

dane1122 commented Apr 11, 2017

avirshup commented Apr 18, 2017 •

edited

		# See http://parmed.github.io/ParmEd/html/index.html for ParmEd units

		def _create_lammps_data(parmedmol):

Interactive simulation #129

Are you sure you want to change the base?

Interactive simulation #129

Conversation

dane1122 commented Feb 1, 2017

avirshup left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dane1122 commented Apr 7, 2017

avirshup left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

avirshup left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

avirshup commented Apr 11, 2017 • edited

avirshup Apr 11, 2017 • edited

Choose a reason for hiding this comment

ktbolt Apr 11, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dane1122 commented Apr 11, 2017

avirshup commented Apr 11, 2017

dane1122 commented Apr 11, 2017

avirshup commented Apr 18, 2017 • edited

avirshup commented Apr 11, 2017 •

edited

avirshup Apr 11, 2017 •

edited

ktbolt Apr 11, 2017 •

edited

avirshup commented Apr 18, 2017 •

edited