Issues with particle extraction in Warp

[Peng Liu]

Dear all,

We encountered some issues when extracting sub-volumes using Warp (Windows version). Our workflow was as follows:

(1) Tilt series alignment with AreTomo2.
(2) Tomogram reconstruction in Warp, contrast inverted, with pixel size 10 Å (original unbinned tomogram pixel size: 1.98 Å/pixel).
(3) Template matching using PyTom match-pick, which yielded ~10k particles per tomogram. Inspection in IMOD suggested the matches were convincing (see screenshot, matched particles in green circles).
(4) Particle extraction in Warp at a smaller pixel size, box size 40 pixels, with a particle diameter of ~10 nm, intended for further processing in Relion.

Here, we encountered problems (see screenshots). When checking individual extracted particles: In IMOD, although the box size was set to 40 Å, it appeared that multiple particles fit within this box; In ChimeraX, the extracted volumes appeared very noisy, possibly containing multiple particles.

We observed similar issues when using GAPSTOP and Dynamo after template matching.

Our questions are:
(1) Could this be due to the extraction parameters we used in Warp?
(2) Or are these simply false positives from template matching, and what we observed is actually expected?
(3) If this issue is related to our old Warp version, are there other tools you would recommend for particle extraction that are compatible with Warp/PyTom/Relion?

In Warp’s “Export sub-tomograms” dialog, there are options such as Volumes vs Image Series, and checkboxes for Normalize input images and Normalize output volumes. Could you advise what these options mean in practice, and which settings are recommended for downstream processing?

Thank you in advance for your help. I would greatly appreciate your advice and experience.

Best regards,

Peng

--

Dr. Peng Liu

Max-Planck-Institut für biophysikalische Chemie
Am Faßberg 11, 37077 Göttingen
Germany

 : +49(0)17635209768
Email: peng...@mpibpc.mpg.de

[Alister Burt]

Hi Peng,

10k particles per tomogram is extremely high and implies a high false positive rate (although I agree your IMOD screenshots implies a reasonable particle density)

With such a high false positive rate most of your particles are likely just noise. You need to fix that.

Pixel size issues can only happen if an incorrect pixel size was entered at some point, check carefully if you still think this is the case.

For everything else, you are now using a very outdated warp installation and we recommend moving to WarpTools.

Particle series are images of the particles extracted from the tilt series which have been CTF corrected - this export used to only be used by warp and a few downstream packages, RELION was more recently updated to work with 2D particle series but the metadata model is different so we now export RELION compatible metadata with these images instead in WarpTools.

Normalization is important in cryo-EM image processing in general and different normalization strategies may make more sense depending on the specifics of your data.

We currently recommend using WarpTools for processing and to generate 2D particle stacks for initial 3D refinement in RELION.

Cheers,

Alister

Sent from mobile - apologies for brevity

On Sep 29, 2025, at 06:54, Peng Liu <penggr...@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 visit https://groups.google.com/d/msgid/warp-em/CALVQGL9ynrOJaxVWnjwPHuf0hYK5LY8bNm60eqwfNwKwu0%3DrsA%40mail.gmail.com.

<IMODindividualparticle.png>

<Warpparticleextraction.png>

<TM-min.png>

<ChimeraXindividualparticle.png>

[Peng Liu]

Hi Alister,

Thank you very much for your helpful suggestions. Our target is relatively small (~10 nm in size) and could be somewhat dense within the lamella, but I agree that 10k particles per tomogram is likely too high. The IMOD image I initially shared was deconvoluted, so the original tomogram used for template matching was probably noisier. We will also double-check the pixel size and other parameters as you suggested.

In the meantime, I noticed that some papers and labs are still using RELION 3 rather than RELION 4/5. If I understand correctly, RELION 3 works with 3D subtomograms, while RELION 5 supports both 3D subtomograms and 2D particle stacks. At this stage, I am not entirely sure which approach (2D vs. 3D) would work better for our case. Since it may take some time for us to set up WarpTools on our Linux system, do you think our current combination of Windows Warp and RELION 3 would still be compatible and usable for the initial processing?

Best regards,

Peng

[Alister Burt]

Hi Peng,

I don't see much in the way of meaningful signal in your tomogram slice/particle - I would suggest critically assessing your tilt series alignments and resulting tomogram quality before charging ahead further.

Warp desktop -> RELION 3 is a workflow that worked well and was nice in a lot of ways but the reality is the ecosystem has moved past both of these solutions (and backwards wrt a nice GUI)... you are of course welcome to continue processing in this way but support for this workflow cannot be provided and you may run into big bad bugs.

Cheers,

Alister

[Peng Liu]

Hi Alister,

Thank you for pointing out the concern regarding the signal. We will carefully reassess the tilt series alignments and tomogram quality. I fully agree that having very good tomograms makes everything else much easier.

It is also very helpful to hear your perspective on the current situation with Warp and RELION. We are still relatively new to RELION 5, but we will hopefully set up the new WarpTools as soon as possible.

Best regards,

Peng

Dr. Peng Liu

Max-Planck-Institut für biophysikalische Chemie
Am Faßberg 11, 37077 Göttingen
Germany

 : +49(0)17635209768
Email: peng...@mpibpc.mpg.de