Re: [chimerax-users] Trajectory computation slowing down overtime

13 Jun 2022

      Hi Shubham,
	Looking at your graph, even the initial time is way too slow, 10+ seconds/frame.  For this kind of analysis, you should be using low level mathematical primitives rather than the graphical axes/planes/centroids, which are intended for interactive use and have a lot of additional overhead, like a triangle mesh for the graphical depiction, color information, etc.
	So the first thing is you will need to get is the atoms that the primitives will based on.  Assuming you know how to select the atoms with a command, you can get them with:

	from chimerax.atomic import selected_atoms
	run(session, "sel some-atom-spec")
	atoms = selected_atoms(session)

	So, the low level primitives are:

Centroid: just an xyz point, which you can get with:
	from chimerax.centroid import centroid
	centroid_pt = centroid(atoms.coords)
	(which is just atoms.coords.mean(0) actually)

Axis: as long as you not measuring distances with an axis, all you need is a direction vector, which is also just an xyz value, which you can get with:
	from numpy.linalg import svd
	center = atoms.coords.mean(0)
	_, vals, vecs = sud(atoms.coords - center, full_matrices=False)
	xyz = vecs[vals.argsort()[-1])

Plane: an origin point and a normal vector, encapsulated as a chimera.geometry.Plane object, e.g.:
	from chimerax.geometry import Plane
	plane = Plane(atoms.coords)

Here are example measurements:

Centroid-centroid distance:
	from chimerax.geometry import distance
	d = distance(xyz1, xyz2)

Centroid-plane distance:
	d = abs(plane.distance(xyz))

Axis-axis angle:
	from chimerax.geometry import angle
	a = angle(xyz1, xyz2)
This returns a value in the range 0-180.  To map that into the 0-90 range:
	if a > 90:
		a = 180 - a

Plane-plane angle:
	from chimerax.geometry import angle
	a = angle(plane1.normal, plane2.normal)
	if a > 90:
		a = 180 - a

--Eric
...
On Jun 13, 2022, at 11:38 AM, Shubham Devesh Ramgoolam via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Hi Eric,
No, I did not.
I tried it now by adding the following code at the end of the first for loop: run(session, ‘close  #1.*’)
It made the issue worse, the computation took longer than before. I believe it is because it takes time to close these models.
(Originally, I tried fixing the issue by closing the entire session and starting a new one. However, it took too much time for chimeraX to close a session, after defining lots of geometric objects.)
Is there a quicker to close these models at the end of each frame?
Shubham
Sent from Mail <https://go.microsoft.com/fwlink/?LinkId=550986> for Windows
From: Eric Pettersen <mailto:pett@cgl.ucsf.edu>
Sent: Monday, June 13, 2022 20:43
To: Shubham Devesh Ramgoolam <mailto:sdramgoolam@uwaterloo.ca>
Cc: chimerax-users@cgl.ucsf.edu <mailto:chimerax-users@cgl.ucsf.edu>
Subject: Re: [chimerax-users] Trajectory computation slowing down overtime
Hi Shubham,
Defining those geometric objects creates models.  Are you closing those models at the end of each frame?
--Eric
Eric Pettersen
UCSF Computer Graphics Lab
On Jun 11, 2022, at 11:21 PM, Shubham Devesh Ramgoolam via ChimeraX-users <chimerax-users@cgl.ucsf.edu <mailto:chimerax-users@cgl.ucsf.edu>> wrote:
I just wanted to specify that I am using the nogui mode of chimeraX for the computation.
Shubham.
<CB8387F6EE204E1BA668D200AE4A00A4.png>
From: Shubham Devesh Ramgoolam
Sent: June 12, 2022 1:38 AM
To: chimerax-users@cgl.ucsf.edu <mailto:chimerax-users@cgl.ucsf.edu> <chimerax-users@cgl.ucsf.edu <mailto:chimerax-users@cgl.ucsf.edu>>
Subject: Trajectory computation slowing down overtime
Hi all,
I hope you are doing well.
I have a Python script for analyzing specific residue pairs for every frame in a trajectory.
It does so, by using 2 for loops:
For frame in list_frames:
               Set current frame to frame
               For pair in specific_residue_pairs:
                              Do computation
The computation involves defining geometric objects(centroids, axes, planes) and performing calculations with them(distance, angles).
The issue here is that as we move further in list_frames, the computation takes longer and longer. There is a somewhat linear increase in computation time, as illustrated below:
<E0E35EA38D4345BE98A480EB08BE8CEC.png>
Is there a way to maintain the same computation time for every frame in the trajectory?
Regards,
Shubham Ramgoolam
Sent from Mail <https://go.microsoft.com/fwlink/?LinkId=550986> for Windows
_______________________________________________
ChimeraX-users mailing list
ChimeraX-users@cgl.ucsf.edu <mailto:ChimeraX-users@cgl.ucsf.edu>
Manage subscription:
https://www.rbvi.ucsf.edu/mailman/listinfo/chimerax-users <https://www.rbvi.ucsf.edu/mailman/listinfo/chimerax-users>
_______________________________________________
ChimeraX-users mailing list
ChimeraX-users@cgl.ucsf.edu <mailto:ChimeraX-users@cgl.ucsf.edu>
Manage subscription:
https://www.rbvi.ucsf.edu/mailman/listinfo/chimerax-users <https://www.rbvi.ucsf.edu/mailman/listinfo/chimerax-users>