Relion resolution issues with Warp tomogram processing

[cla...@gmail.com]

Hi all,

I've been working my way through the tomogram processing pipeline using Warp. I have several tilt series collected on a K3 (2.265 A superres), I binned them to 4.53 A during Warp processing, tiltaligned using Dynamo's automated process and reconstructed with a pixel size of 18.12 A so I could pick particles. I then extracted the (~1200) subvolumes in Warp using pixel sizes of 4.53, 9.06 and 18.12 A. My initial plan was to check at 18.12 A and confirm that everything had worked, then work my way to the smallest pixel size with subsequent refinement steps in Relion, then finally switching to M.

When I process in Relion, I can see the map improving with each iteration but for some reason the resolution remains stuck at 58 A / 39 A (for 18.12 A and 9.06 A respectively) and the estimated angular accuracy stays at 0.05 degrees no matter what. Applying a mask around my particle increases the resolution from 39 A to 38 A.

I've previously processed the data using the Relion STA pipeline (i.e. this) and achieved a resolution of ~ 20 A so this isn't a limitation of the dataset from what I can tell, although in the case of the 20 A map I did all the processing through IMOD instead of the automated tiltalign in Dynamo.

I also noticed that the 3D CTFs are a strange shape, the same as mentioned in this post here but it wasn't clear if this is normal or something I should be worried about.

I haven't used any unusual settings in either Warp or Relion as I've been vaguely following the tutorial here with adjustments as necessary for my data. Have these issues been encountered by anyone else?

Cheers,

Mat

[Pranav Shah]

Hi Mat,

I would start by assessing if the tilt alignments are locked in. Since you did the alignments in IMOD how do the shifts in the xf file compare against the shifts calculated in dynamo?

Can you re-run your warp-relion-M pipeline by just substituting the Dynamo xf files with the ones from IMOD?

Regarding the CTF files -  that's totally normal.

[cla...@gmail.com]

Hi all,

I've got a little update on this, although unfortunately I don't think I'm much closer to solving the problem.

I followed Pranav's advice and reconstructed the tomograms using IMOD's shifts, then repicked the particles (as the tomograms were different sizes). I re-extracted them in Warp and loaded them in Relion and still ended up with the same issue. Comparing the IMOD/Relion map to the Warp/IMOD/Relion map shows that the latter is lower resolution. Alister suggested checking the FSC masked vs corrected from a Relion postprocess, which I've also attached. The curve increases slightly at the higher resolutions and I'm not sure why, but I would guess that this is a symptom rather than a cause.

On top of that, we've been trying a different dataset (on a Krios/Falcon 4 this time) and experienced a similar issue. It's still early days with the processing but we still see the 0.05 degree angular accuracy issue and the resolution stays pinned around the lowpassed map. This second dataset hasn't been processed without Warp but we have processed a dataset of the same sample on a 200 kV microscope and the Warp-processed map looks noisier.

As nobody else has mentioned similar issues on this forum (and it's happened with two datasets) I'm assuming it's something I'm doing wrong and I have three possible ideas of what could affect it. Apologies for a general question, but do you think that a "wrong" amplitude contrast value could negatively affect this? I've always used 0.07 but could using 0.1 improve anything? I don't think I've ever seen a discussion on when to change this value so I don't know if it can be beneficial for some data. Secondly, with IMOD alignment, is Warp capable of using the minimisation parameters and local alignment parameters settings from the fine alignment tab? I'm less convinced that this is the problem since I had the same issue with Dynamo's automated tilt alignment. Finally, could the wrong particle diameter in Warp's subvolume export cause problems depending on particle shape? Neither of our samples are near spherical but I've been using the longest side for the diameter and wondered if there are knock-on issues more than Dimitry's mention of incorrect normalisation warnings in Relion.

I'd very much appreciate any further suggestions if you have any!

Cheers,

Mat

[Mathew McLaren]

Hi Pranav,

It's definitely not the wrong tilt handedness, I was careful to make sure that I checked it previously. I've attached a sample CTF curve from one of my tomograms, I do wonder if I could play with the settings to make it more accurate.

Cheers,

Mat

On Mon, 10 May 2021 at 19:02, Pranav Shah <Prana...@strubi.ox.ac.uk> wrote:

Hi Mat,
I am starting to suspect that you most likely have wrong tilt
handedness.Have you run the tilt handedness check in Warp? If not, run
the tilt-handedness check
and re-export the subtomos and report back.

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

On Mon, May 10, 2021 at 4:19 PM cla...@gmail.com <cla...@gmail.com> wrote:
>
> Hi all,
>
> I've got a little update on this, although unfortunately I don't think I'm much closer to solving the problem.
> I followed Pranav's advice and reconstructed the tomograms using IMOD's shifts, then repicked the particles (as the tomograms were different sizes). I re-extracted them in Warp and loaded them in Relion and still ended up with the same issue. Comparing the IMOD/Relion map to the Warp/IMOD/Relion map shows that the latter is lower resolution. Alister suggested checking the FSC masked vs corrected from a Relion postprocess, which I've also attached. The curve increases slightly at the higher resolutions and I'm not sure why, but I would guess that this is a symptom rather than a cause.
> On top of that, we've been trying a different dataset (on a Krios/Falcon 4 this time) and experienced a similar issue. It's still early days with the processing but we still see the 0.05 degree angular accuracy issue and the resolution stays pinned around the lowpassed map. This second dataset hasn't been processed without Warp but we have processed a dataset of the same sample on a 200 kV microscope and the Warp-processed map looks noisier.
>
> As nobody else has mentioned similar issues on this forum (and it's happened with two datasets) I'm assuming it's something I'm doing wrong and I have three possible ideas of what could affect it. Apologies for a general question, but do you think that a "wrong" amplitude contrast value could negatively affect this? I've always used 0.07 but could using 0.1 improve anything? I don't think I've ever seen a discussion on when to change this value so I don't know if it can be beneficial for some data. Secondly, with IMOD alignment, is Warp capable of using the minimisation parameters and local alignment parameters settings from the fine alignment tab? I'm less convinced that this is the problem since I had the same issue with Dynamo's automated tilt alignment. Finally, could the wrong particle diameter in Warp's subvolume export cause problems depending on particle shape? Neither of our samples are near spherical but I've been using the longest side for the diameter and wondered if there are knock-on issues more than Dimitry's mention of incorrect normalisation warnings in Relion.
> I'd very much appreciate any further suggestions if you have any!
>
> Cheers,
>
> Mat
>
> On Thursday, 25 February 2021 at 12:43:16 UTC p.sha...@gmail.com wrote:
>>
>> Hi Mat,
>>
>> I would start by assessing if the tilt alignments are locked in. Since you did the alignments in IMOD how do the shifts in the xf file compare against the shifts calculated in dynamo?
>> Can you re-run your warp-relion-M pipeline by just substituting the Dynamo xf files with the ones from IMOD?
>>
>> Regarding the CTF files -  that's totally normal.
>>
>> On Thursday, February 25, 2021 at 12:26:00 PM UTC cla...@gmail.com wrote:
>>>
>>> Hi all,
>>>
>>> I've been working my way through the tomogram processing pipeline using Warp. I have several tilt series collected on a K3 (2.265 A superres), I binned them to 4.53 A during Warp processing, tiltaligned using Dynamo's automated process and reconstructed with a pixel size of 18.12 A so I could pick particles. I then extracted the (~1200) subvolumes in Warp using pixel sizes of 4.53, 9.06 and 18.12 A. My initial plan was to check at 18.12 A and confirm that everything had worked, then work my way to the smallest pixel size with subsequent refinement steps in Relion, then finally switching to M.
>>> When I process in Relion, I can see the map improving with each iteration but for some reason the resolution remains stuck at 58 A / 39 A (for 18.12 A and 9.06 A respectively) and the estimated angular accuracy stays at 0.05 degrees no matter what. Applying a mask around my particle increases the resolution from 39 A to 38 A.
>>>
>>> I've previously processed the data using the Relion STA pipeline (i.e. this) and achieved a resolution of ~ 20 A so this isn't a limitation of the dataset from what I can tell, although in the case of the 20 A map I did all the processing through IMOD instead of the automated tiltalign in Dynamo.
>>>
>>> I also noticed that the 3D CTFs are a strange shape, the same as mentioned in this post here but it wasn't clear if this is normal or something I should be worried about.
>>>
>>> I haven't used any unusual settings in either Warp or Relion as I've been vaguely following the tutorial here with adjustments as necessary for my data. Have these issues been encountered by anyone else?
>>>
>>> Cheers,
>>>
>>> Mat
>

> --
> You received this message because you are subscribed to a topic in the Google Groups "Warp" group.
> To unsubscribe from this topic, visit https://groups.google.com/d/topic/warp-em/eV5xrn2GllM/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to warp-em+u...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/warp-em/1e235f74-978c-40c5-9a44-44ad7cdcc52bn%40googlegroups.com.

[Mathew McLaren]

Hi Pranav,

I'm relatively certain I did but I processed the first tomogram, checked tilt handedness and then processed the rest so at worst only one will be incorrect (which for one dataset at least had no particles anyway). Anyway, I reran the processing as a sanity check, re-extracted the particles and tried processing in Relion and still have the same resolution/angular accuracy issues.

Mat

On Mon, 10 May 2021 at 21:09, Pranav Shah <Prana...@strubi.ox.ac.uk> wrote:

Did you re-estimate the CTF afterwards? 

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 on the web visit https://groups.google.com/d/msgid/warp-em/CAA%3DxVZvYxeAJiN6PPhWypj0UkiKC26NNKC5rjiMzQh%2B81s4t3A%40mail.gmail.com.

--

[Alister Burt]

Hi Mat,

Nothing exceptionally useful to add unfortunately, only that I and others have seen the angular accuracy reported in RELION 3.X during refinement of Warp particles drop extremely low whilst reporting low resolutions for the map. This obviously affects the regularisation in RELION meaning it never has a chance to access the higher res info and was one reason I performed some alignments in Dynamo for the chemosensory array processing described in this preprint https://www.biorxiv.org/content/10.1101/2021.01.31.428990v1. I’m not sure exactly what’s going on and didn’t see the same behaviour (or at least not to the same extent) with the HIV dataset. Part of my motivation for reprocessing the Briggs data was frustration with exactly this issue, I might suggest doing the same to convince yourself what you’re doing should work, it doesn’t take too long.

Are you performing sanity checks at every step, making sure that reconstructing your particles prior to refinement gives a meaningful average? If you have a set of poses which give a relatively good map it might be worth dropping into M directly

Cheers and good luck! 

Alister

To view this discussion on the web visit https://groups.google.com/d/msgid/warp-em/CAA%3DxVZshr36aZ-qEi0Q95T7ZtLhCcFXmmGKn0%3DWmVFn7ikV9cw%40mail.gmail.com.

[Pranav Shah]

Hi Mat,
I am starting to suspect that you most likely have wrong tilt
handedness.Have you run the tilt handedness check in Warp? If not, run
the tilt-handedness check
and re-export the subtomos and report back.

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

On Mon, May 10, 2021 at 4:19 PM cla...@gmail.com <cla...@gmail.com> wrote:
>
> Hi all,
>
> I've got a little update on this, although unfortunately I don't think I'm much closer to solving the problem.
> I followed Pranav's advice and reconstructed the tomograms using IMOD's shifts, then repicked the particles (as the tomograms were different sizes). I re-extracted them in Warp and loaded them in Relion and still ended up with the same issue. Comparing the IMOD/Relion map to the Warp/IMOD/Relion map shows that the latter is lower resolution. Alister suggested checking the FSC masked vs corrected from a Relion postprocess, which I've also attached. The curve increases slightly at the higher resolutions and I'm not sure why, but I would guess that this is a symptom rather than a cause.
> On top of that, we've been trying a different dataset (on a Krios/Falcon 4 this time) and experienced a similar issue. It's still early days with the processing but we still see the 0.05 degree angular accuracy issue and the resolution stays pinned around the lowpassed map. This second dataset hasn't been processed without Warp but we have processed a dataset of the same sample on a 200 kV microscope and the Warp-processed map looks noisier.
>
> As nobody else has mentioned similar issues on this forum (and it's happened with two datasets) I'm assuming it's something I'm doing wrong and I have three possible ideas of what could affect it. Apologies for a general question, but do you think that a "wrong" amplitude contrast value could negatively affect this? I've always used 0.07 but could using 0.1 improve anything? I don't think I've ever seen a discussion on when to change this value so I don't know if it can be beneficial for some data. Secondly, with IMOD alignment, is Warp capable of using the minimisation parameters and local alignment parameters settings from the fine alignment tab? I'm less convinced that this is the problem since I had the same issue with Dynamo's automated tilt alignment. Finally, could the wrong particle diameter in Warp's subvolume export cause problems depending on particle shape? Neither of our samples are near spherical but I've been using the longest side for the diameter and wondered if there are knock-on issues more than Dimitry's mention of incorrect normalisation warnings in Relion.
> I'd very much appreciate any further suggestions if you have any!
>
> Cheers,
>
> Mat
>
> On Thursday, 25 February 2021 at 12:43:16 UTC p.sha...@gmail.com wrote:
>>
>> Hi Mat,
>>
>> I would start by assessing if the tilt alignments are locked in. Since you did the alignments in IMOD how do the shifts in the xf file compare against the shifts calculated in dynamo?
>> Can you re-run your warp-relion-M pipeline by just substituting the Dynamo xf files with the ones from IMOD?
>>
>> Regarding the CTF files - that's totally normal.
>>
>> On Thursday, February 25, 2021 at 12:26:00 PM UTC cla...@gmail.com wrote:
>>>
>>> Hi all,
>>>
>>> I've been working my way through the tomogram processing pipeline using Warp. I have several tilt series collected on a K3 (2.265 A superres), I binned them to 4.53 A during Warp processing, tiltaligned using Dynamo's automated process and reconstructed with a pixel size of 18.12 A so I could pick particles. I then extracted the (~1200) subvolumes in Warp using pixel sizes of 4.53, 9.06 and 18.12 A. My initial plan was to check at 18.12 A and confirm that everything had worked, then work my way to the smallest pixel size with subsequent refinement steps in Relion, then finally switching to M.
>>> When I process in Relion, I can see the map improving with each iteration but for some reason the resolution remains stuck at 58 A / 39 A (for 18.12 A and 9.06 A respectively) and the estimated angular accuracy stays at 0.05 degrees no matter what. Applying a mask around my particle increases the resolution from 39 A to 38 A.
>>>
>>> I've previously processed the data using the Relion STA pipeline (i.e. this) and achieved a resolution of ~ 20 A so this isn't a limitation of the dataset from what I can tell, although in the case of the 20 A map I did all the processing through IMOD instead of the automated tiltalign in Dynamo.
>>>
>>> I also noticed that the 3D CTFs are a strange shape, the same as mentioned in this post here but it wasn't clear if this is normal or something I should be worried about.
>>>
>>> I haven't used any unusual settings in either Warp or Relion as I've been vaguely following the tutorial here with adjustments as necessary for my data. Have these issues been encountered by anyone else?
>>>
>>> Cheers,
>>>
>>> Mat
>

[Pranav Shah]

Did you re-estimate the CTF afterwards? 

On Mon, 10 May 2021 at 20:42, Mathew McLaren <cla...@gmail.com> wrote:

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 on the web visit https://groups.google.com/d/msgid/warp-em/CAA%3DxVZvYxeAJiN6PPhWypj0UkiKC26NNKC5rjiMzQh%2B81s4t3A%40mail.gmail.com.

--

[Pranav Shah]

I am sorry to hear that Mat, I must admit that in the absence of
further information, it is hard to imagine what could be going wrong
at this point...

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 on the web visit https://groups.google.com/d/msgid/warp-em/CAA%3DxVZshr36aZ-qEi0Q95T7ZtLhCcFXmmGKn0%3DWmVFn7ikV9cw%40mail.gmail.com.