PDB dimensionality reduction improvements #115

mms29 · 2022-04-21T15:27:17Z

Add a trajectory viewer to the PDB dimensionality reduction protocol.
Add the possibility to input PDBs from scipion object SetOfPDBs and from DCD trajectory files
Reduction methods are limited to sklearn PCA (because xmipp methods were too slow)
Replace the private reading of PDB by a PDB handler (protocols/utilities/pdb_handler.py) called ContinuousFlexPDBHandler designed to read/write/align/compare PDBs.

continuousflex/protocols/protocol_pdb_dimred.py

MohamadHarastani · 2022-04-22T04:32:35Z

continuousflex/protocols/protocol_pdb_dimred.py

@@ -64,7 +51,7 @@ def _defineParams(self, form):
        form.addSection(label='Input')
        form.addParam('pdbSource', EnumParam, default=0,
                      label='Source of PDBs',
-                      choices=['Used for subtomogram synthesis', 'File pattern'],
+                      choices=['Used for subtomogram synthesis', 'File pattern', 'Object', 'Trajectory Files'],
                      help='Use the file pattern as file location with /*.pdb')
        form.addParam('pdbs', params.PointerParam, pointerClass='FlexProtSynthesizeSubtomo',


I think we can add another pointerClass and let this work for synthesized images not only subtomograms right?

Yes it would be good, we can put that as an issue for later as I won't have time to do it right away

Lets first try to do that:
pointerClass='FlexProtSynthesizeSubtomo, FlexProtSynthesizeImages',

Okay, I understand now what you said. Actually, wouldn't it be even better to output a set of pdbs from these synthesize protocols?
What I understood at the beginning was to allow synthesizing from the same input that pdb dimred (pdb object, dcd trajectory, etc.) That would be also helpful (e.g. generate a MD simulation, then synthesize images/subtomograms)

It is a good idea to output the set of PDBs from the synthesize protocols. An exception can be made if the user does not have a lot of space, and we can finally delete them to save disk space.

MohamadHarastani · 2022-04-22T04:50:13Z

continuousflex/protocols/protocol_pdb_dimred.py

+                      label="Alignement Reference PDB",
+                      help='Reference PDB to align the PDBs with')
+        form.addParam('matchingType', params.EnumParam, label="Match structures ?", default=0,
+                      choices=['Both structures are the same', 'Match chain name/residue num/atom name',


"Both" -> "All"
"same" -> ?
For same, we need to find a smart word. Maybe something like "have 1-1 correspondence between the atoms in the same order"?

What about "All structures are matching", and in the help I say "Method to find atomic coordinates correspondence between the trajectory coordinates and the reference PDB. The method will select the matching atoms and sort them in the corresponding order. If the structures in the files are already matching, choose "All structures are matching"

MohamadHarastani · 2022-04-22T05:06:33Z

Good job in the a pdb handler, I hope it is not buggy haha!
I tested PDB dimred on a set of PDBs coming from synthesized subtomograms (AK, linear relationship, 100 volumes). The conformational space is not as expected. I saw two lines instead of one.
The animations were not correct either.
We can get rid of the Error bar. I know it is annoying to use the viewers with having to set the weight etc, but I solved that before in TomoFlow viewers (they don't have a color bar on the right).
Is there a reason we are not showing the conformational space without trajectories toots? I think it is not needed since there is only trajectories, so good job on that.

mms29 · 2022-04-22T13:02:52Z

Good job in the a pdb handler, I hope it is not buggy haha!

The way I read and write PDB line is copied from biopython. Also I used this reader/writer for a long time on many PDBs so It should be fine (It should ...)

I tested PDB dimred on a set of PDBs coming from synthesized subtomograms (AK, linear relationship, 100 volumes). The conformational space is not as expected. I saw two lines instead of one. The animations were not correct either.

You are right, I did the same and I have these two lines. It seems that one PC is good and not the other, I will investigate

We can get rid of the Error bar. I know it is annoying to use the viewers with having to set the weight etc, but I solved that before in TomoFlow viewers (they don't have a color bar on the right).

That is great, I wanted to do that but I faced that I need to change different files so I left that for later. If you have something that does it please share.

Is there a reason we are not showing the conformational space without trajectories toots? I think it is not needed since there is only trajectories, so good job on that.

I removed it since it seemed doing the same thing as the trajectory tool

mms29 · 2022-04-22T13:34:50Z

I tested PDB dimred on a set of PDBs coming from synthesized subtomograms (AK, linear relationship, 100 volumes). The conformational space is not as expected. I saw two lines instead of one. The animations were not correct either.

Actually, I did several tests but everything looks fine, I observe these two lines only on C-alpha AK and actually the PC of these two lines is extremely small compare to the first PC. When looking the trajectory, everything looks fine. I noticed that I don't have these two lines when generating the PDBs myself, so it might come from the synthesize protocol

MohamadHarastani · 2022-04-22T13:39:06Z

Hmm.. interesting. What about the animation? it is flashing for me and is not showing the movement

mms29 · 2022-04-22T13:44:00Z

Is it a visualization problem ? Or the trajectory does not correspond ?

MohamadHarastani · 2022-04-23T06:31:03Z

It looks like that (few sparse points and the structure is hidden)

MohamadHarastani · 2022-04-23T06:39:43Z

I have just tried to give it a file pattern and it does not allow me to execute. I think it's a problem of "allowsNull". Are you sure you don't have some unpushed commits? lol

mms29 · 2022-04-25T06:43:02Z

Okay that is very strange. Can you share the PDBs you used for this plot ?

MohamadHarastani · 2022-04-26T12:48:28Z

Okay that is very strange. Can you share the PDBs you used for this plot ?

I generated them using the volume synthesis protocol, I can send them of course. I commented up that I could not give a file pattern as input, can you?

MohamadHarastani · 2022-05-23T16:58:47Z

Hey Remi.. I tested again and didn't notice any problem. Let me know if you are working on something here or we can merge these changes.

mms29 added 2 commits April 21, 2022 17:13

pdb dimred, PCA, trajectory viewer

aec481f

pdb dimred, PCA, trajectory viewer

2e5946b

mms29 requested a review from MohamadHarastani April 21, 2022 15:27

MohamadHarastani reviewed Apr 22, 2022

View reviewed changes

MohamadHarastani linked an issue Apr 26, 2022 that may be closed by this pull request

Adding trajectory animations to PDB dimred protocol #95

Closed

write pdbs of traj files

21329fd

MohamadHarastani approved these changes May 23, 2022

View reviewed changes

pull request updates

9b0579a

mms29 merged commit 8ae9e74 into devel May 30, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

PDB dimensionality reduction improvements #115

PDB dimensionality reduction improvements #115

mms29 commented Apr 21, 2022

MohamadHarastani Apr 22, 2022

mms29 Apr 22, 2022

MohamadHarastani Apr 22, 2022 •

edited

Loading

mms29 Apr 22, 2022

MohamadHarastani May 23, 2022

MohamadHarastani Apr 22, 2022

mms29 Apr 22, 2022

MohamadHarastani Apr 22, 2022

MohamadHarastani commented Apr 22, 2022

mms29 commented Apr 22, 2022

mms29 commented Apr 22, 2022

MohamadHarastani commented Apr 22, 2022

mms29 commented Apr 22, 2022

MohamadHarastani commented Apr 23, 2022

MohamadHarastani commented Apr 23, 2022

mms29 commented Apr 25, 2022

MohamadHarastani commented Apr 26, 2022

MohamadHarastani commented May 23, 2022

PDB dimensionality reduction improvements #115

PDB dimensionality reduction improvements #115

Conversation

mms29 commented Apr 21, 2022

MohamadHarastani Apr 22, 2022

Choose a reason for hiding this comment

mms29 Apr 22, 2022

Choose a reason for hiding this comment

MohamadHarastani Apr 22, 2022 • edited Loading

Choose a reason for hiding this comment

mms29 Apr 22, 2022

Choose a reason for hiding this comment

MohamadHarastani May 23, 2022

Choose a reason for hiding this comment

MohamadHarastani Apr 22, 2022

Choose a reason for hiding this comment

mms29 Apr 22, 2022

Choose a reason for hiding this comment

MohamadHarastani Apr 22, 2022

Choose a reason for hiding this comment

MohamadHarastani commented Apr 22, 2022

mms29 commented Apr 22, 2022

mms29 commented Apr 22, 2022

MohamadHarastani commented Apr 22, 2022

mms29 commented Apr 22, 2022

MohamadHarastani commented Apr 23, 2022

MohamadHarastani commented Apr 23, 2022

mms29 commented Apr 25, 2022

MohamadHarastani commented Apr 26, 2022

MohamadHarastani commented May 23, 2022

MohamadHarastani Apr 22, 2022 •

edited

Loading