Imod vs warp reconstruction - tilt series alignments import query

[Miles Graham]

Hi all, 

I'm hoping there is something simple to explain this, but I've noticed a difference in my imod vs warp reconstructions of the same tomogram. I performed tilt series alignment externally (finally with imod using fiducial positions from dynamo). I then imported these alignments into warp. I'll provide the commands and relevant files at the end of this message. 

My Imod reconstructions look lovely (gold beads spherical), whereas my warp reconstructions appear slightly stretched. Is there a reason for this? Might it represent an issue with my tilt series alignments import and could it affect downstream averaging? 

I'm sure there is something I haven't considered yet, but time is especially precious at the moment, so decided to ask on here!

Best wishes and many thanks for your time,

Miles

Commands : 

WarpTools ts_import_alignments --settings tiltseries.settings --alignments tiltseries/tiltstack --alignment_angpix 1.864 --input_data input_list_2.txt

WarpTools ts_defocus_hand --settings tiltseries.settings --set_flip --input_data input_list_2.txt

WarpTools ts_ctf --settings tiltseries.settings --range_high 7 --defocus_max 8 --input_data input_list_2.txt

WarpTools ts_reconstruct --settings tiltseries.settings --angpix 5.592 --perdevice 1 --dont_invert --halfmap_frames --input_data input_list_2.txt

[Hamidreza Rahmani]

Dear Miles,

I might be wrong here, but one thing is WARP does a ctf correction before reconstruction where etomo does not. This in our case helps us in seeing the membranes better but there are artifacts. 

Also, seeing the first 6 tilts being "exactly" as the rawtlt makes me think you skipped them in fine alignment? although to me all the angles look like they did not change in alignment? can you send your taSolution.log ?

Best,

Hamid

--
You received this message because you are subscribed to the Google Groups "Warp" group.
To unsubscribe from this group and stop receiving emails from it, send an email to warp-em+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/warp-em/fbeb009f-5866-4d33-9098-a7cb6b1a952fn%40googlegroups.com.

[Miles Graham]

Dear Hamid, 

Thank you for your reply :) 

That is true re. Ctf correction, but I am unsure whether this stretching is expected with that...

Re. the imod alignment process, I fixed a lot of parameters during tiltalign (see start of tiltalign.stdout), hence the static tilt angles. Regardless the information you are looking for can be found in the middle of the tiltalign.stdout provided on the original post. My alignment scripted, so no separate taSolution.log. 

Many thanks for your time and interest,

Miles

[Alister Burt]

Hey Miles,

Cool to see someone using dynamo for fiducial based tsa - I still think it's the best algorithm out there for running that in an automated way.

Could you outline your whole workflow/process for going between warptools -> dynamo -> imod -> warptools? 

Did any of this processing occur with desktop warp/imod before warptools was a thing?

There are some nuances around the handling of tilt angles that may be relevant, if you want to quickly test if this is the thing flip your tilt angles and try to reconstruct again

Cheers,

Alister

To view this discussion visit https://groups.google.com/d/msgid/warp-em/d0f7a999-e043-40d0-9c11-3da01d5a7f44n%40googlegroups.com.

[Pranav Shah]

Hi Alister,

I think chime in a bit - the proc is taking place with warptools, no desktop warp is being used.

Will let miles fill in the blanks.

Best,
Pranav
--
Pranav Shah
Postdoctoral Research Fellow.

Division of Structural Biology,
Wellcome Trust Centre for Human Genetics,
University of Oxford,
Roosevelt Drive, Oxford OX3 7BN,
UK

To view this discussion visit https://groups.google.com/d/msgid/warp-em/CAOt3kHrAevDqXrjAZxMtwY4yYQXXax0tBF0g4Xz%3D4H%3DT139BVA%40mail.gmail.com.

[Miles Graham]

Hi Alister, 

Thanks for your reply! Unfortunately my detailed reply just now hasn't sent. So here we go again... 

First an outline of the pipeline : 

My processing has all been done using WarpTools dev36. 

1/ use ts_stack to generate stacks for alignment

2/ scripted dynamo automated tsa to run on all tilt series

3/ Then ran imod on the IMOD fiducial positions (inspired by your autoalign_dynamo).

4/ Then ran ts_import_alignments

For your suggestion of flipping the tilt angles - it wasn't obvious to me what stage to do this, so I've tried a couple of different things which have resulted in either the same result or garbage. Just for completeness I'll put my attempts below.. Please do point out the flaws if I've run it incorrectly :) 

I'll sign off here. Many thanks for your time - it is very valued as always.

Best wishes,

Miles

Run 1 (a simple starting point) : 

WarpTools ts_import --mdocs mdocs --frameseries frameseries --tilt_exposure 3.3 --output tomostar_dynamo27 --min_intensity 0.2 --override_axis 84.7 --pattern map27_ts_002*.mdoc

WarpTools ts_stack --settings tiltseries27.settings --angpix 1.864 --output_processing tiltseries27

dynamo alignment (as above)

WarpTools ts_import_alignments --settings tiltseries27.settings --alignments tiltseries27/tiltstack --alignment_angpix 1.864 --input_data tomostar_dynamo27/map27_ts_002.tomostar

No defocus handedness flip.

Reconstruction shows the same distortion as my orginal post. 

Run 2 : 

Flipped the order of the tlt and xf files, reimported and reconstructed. 

Resulted in total garbage

Run 3 : 

Reran Run 1 to get back to a sensible start point. Then : 

WarpTools ts_import --mdocs mdocs --frameseries frameseries --tilt_exposure 3.3 --dont_invert --output tomostar_dynamo27 --min_intensity 0.2 --override_axis 84.7 --pattern map27_ts_002*.mdoc

Repeated tilt series alignment import (same tlt and xf file as run 1)

Reconstruction results in the same distortion. 

[Alister Burt]

Hi Miles,

Sorry not immediately sure what to suggest, how is what you’ve done different to autoalign_dynamo? I don’t remember having any issues with alignments done this way… good luck figuring it out!

Cheers,

Alister

Sent from mobile - apologies for brevity

On Nov 2, 2025, at 13:43, Miles Graham <miles.g...@gmail.com> wrote:

Hi Alister, 

To view this discussion visit https://groups.google.com/d/msgid/warp-em/100b2563-9add-4ba5-9a93-02a9e5e83cedn%40googlegroups.com.

[Miles Graham]

Hi Alister,

No worries. I'm not sure it's related to my alignment method, since I just ran the standard pipeline for this tomogram just now and saw the same differences in reconstruction (spherical beads when I make imod reconstructions (no change in alignments) and slightly stretched in warp (no different to the previous images).

Perhaps there is no issue with alignments, but some difference in the Warp reconstruction method? 

The commands I used are below. 

Best wishes,

Miles

WarpTools ts_import --mdocs mdocs --frameseries frameseries --tilt_exposure 3.3 --dont_invert --output tomostar_dynamo27 --min_intensity 0.2 --override_axis 84.7 --pattern map27_ts_002*.mdoc

WarpTools create_settings --output tiltseries27.settings --folder_processing tiltseries27 --folder_data tomostar_dynamo27 --extension "*.tomostar" --angpix 0.932 --gain_path 20250617_154651_EER_GainReference.gain --exposure 3.3 --tomo_dimensions 4096x4096x3000

WarpTools ts_etomo_fiducials --settings tiltseries27.settings --angpix 1.864 --initial_axis 84.7 --fiducial_size 6 --n_beads_target 20 --min_fov 0.2

Defocus handedness flip as recommended by the check

WarpTools ts_ctf --settings tiltseries27.settings --range_high 7 --defocus_max 8 --input_data tomostar_dynamo27/map27_ts_002.tomostar WarpTools ts_reconstruct --settings tiltseries27.settings --angpix 5.59 --perdevice 1 --dont_invert --input_data tomostar_dynamo27/map27_ts_002.tomostar

For note, I am making my reconstruction as follows : 

tilt -input map27_ts_002_ali.mrc -output bin4tomogram_autoamted_imod_alignment.mrc -TILTFILE map27_ts_002.tlt -THICKNESS 400 -IMAGEBINNED 3 -rotate -90

[Alister Burt]

Hi Miles,

This doesn’t match our experience, we like crispy tomograms 😂 can you run the same in a fresh directly, no existing outputs and confirm the observation?

Cheers,

Alister

Sent from mobile - apologies for brevity

To view this discussion visit https://groups.google.com/d/msgid/warp-em/85d6ecc4-0058-4164-8168-5527b61e597cn%40googlegroups.com.

[Miles Graham]

Hi Alister,

I'm sure! I'm really scratching my head too. 

Following 'rm -R tiltseries27 tomostar_dynamo27/ tiltseries27.settings',  I ran as above, but still the same result. 

The tomograms are perfectly interpretable, but I'm just concerned something is off here that would affect the quality of averaging. 

Best wishes,

Miles

[Alister Burt]

Hi Miles,

Thanks for the quick back and forth - still not obvious to me why you would be getting weird results and yes any poor alignments in tomograms will affect downstream averaging - two questions:

- what happens if you re-run alignments at e.g. 10apx?

- is the stretching you see particularly visible in tomograms reconstructed at 10apx?

Cheers,

Alister

Sent from mobile - apologies for brevity

To view this discussion visit https://groups.google.com/d/msgid/warp-em/380bc6b9-5797-4ff6-ae4d-0bd285545e18n%40googlegroups.com.

[Miles Graham]

Hi Alister, 

No worries, keen to get to the bottom of this as there is no point moving on with averaging if this isn't right. 

Rerunning the standard etomo alignments (as previous) at 10 apx and then comparing imod vs warp reconstructions (also 10apx) of the same fiducial result in the following. Still a clear difference (see attached images) 

And yes, the stretching is very noticeable at 10apx reconstructions, even with the original alignments. 

Best wishes,

Miles

[Ricardo Righetto]

Hi Miles,

No direct answers but if I understand correctly, what you want is to reconstruct tomos in WarpTools that look exactly the same as what you get from IMOD's tilt, right?

(except for the differences in filtering/weighting of the data of course, CTF correction also, but the tomos should still pretty much "match" each other)

To me, it looks like WarpTools somehow is not interpreting the TS alignment information (from the .xf file) properly. Why that is, I have no idea. Do the tomos look similar overall? I think I just saw zoom-ins of the beads, but what about the rest of the FOV?

If you look at the FT of the XZ plane, do the tilts match in both tomos? (to make sure the .tlt file is being read in correctly)

The other possible culprit would be a blatant difference in the radial filtering (e.g. to downweight the lower frequencies, which are overrepresented), but that would be weird. And if that was the only problem, the two reconstructions would still match well.

Anyway, I would investigate along these lines.

Hope this helps!

Cheers,

--
Ricardo Diogo Righetto

To view this discussion visit https://groups.google.com/d/msgid/warp-em/0bd10099-d104-42f1-8029-2ab1fdf0f5d0n%40googlegroups.com.

[Pranav Shah]

@miles

Could you check in the tomo xml file if the tilt axis angle and the physical angstrom shifts match those in the imod xf files? Note xf records shifts in px so to convert you need to mult by pali px size.

Best,
Pranav
--
Pranav Shah
Postdoctoral Research Fellow.

Division of Structural Biology,
Wellcome Trust Centre for Human Genetics,
University of Oxford,
Roosevelt Drive, Oxford OX3 7BN,
UK

To view this discussion visit https://groups.google.com/d/msgid/warp-em/CAB7-Rrtv9Qc9kVnr1OvTq_LOr4%3DPHvmWEn7SoSMs12KcdqJpfQ%40mail.gmail.com.

[Miles Graham]

Hi all, 

Thanks for the replies. I'm still fiddling around to try and understand the issue (actually now downloading some EMPIAR data to see if I can recreate the discrepancy). 

In the meantime, I'd find it very helpful to understand why there are differences between the xf files of imod and warp xml files. I've attached a table to show the values for each (converted to angstrom)

The XF X shift roughly matches the XML's AxisOffsetY (with sign flipped), but not exactly. 

For the XML's AxisOffsetX I can't figure out where those values have come from. 

I'm curious what Warp is doing under the hood and whether there is some convention I'm missing? 

Ricardo - thanks for your suggestions! I notice a very small shift in field of view between the two tomograms, but very minor. Again though, curious to understand the source of that, but I reckon the answer lies with the above? I looked at the FT of the XZ plane and the tilts seem to match, or at least weren't obviously different. All your thoughts are much appreciated :) 

Best wishes,

Miles

[Alister Burt]

Hi,

Semantics of what the values mean are completely different in each case - in IMOD they’re shifts applied after rotation in warp they’re shifts applied before rotation

Cheers,

Alister

Sent from mobile - apologies for brevity

To view this discussion visit https://groups.google.com/d/msgid/warp-em/ea61230b-4125-45a2-804a-76fd5923b683n%40googlegroups.com.

<xf_vs_xml.pdf>