Refinement in MCore - Choosing parameters

[Paris Watson]

Hi, 

I am currently wanting trying to refine in M with particles that went to 8A in relion. 

I was just wondering if there were any guidelines on refinement in M and what parameters to use/try and if there is a particular order? I wasnt certain that the parameters in the tutorial were good for every case. 

Thank you for any help you can provide to get started!!

Paris

[Alister Burt]

Hey Paris,

Great question - the need for parameter selection is a weakness (and strength!) of M

General advice is to 
- make sure your particle set is as clean as possible before going into M

- understand what each parameter group in M represents and how they are parametrized

- start conservatively (less parameters == less chance to overfit with weak/noisy data)

- progressively increase set of parameters used until map stops improving

Read the M paper to understand the above

Specific protocol:

- 1x1 image warp + particle poses only, repeat until resolution stops improving

- add defocus (with exhaustive search in first run) if map is <7A or so, repeat until resolution stops improving
- increase spatial resolution of image warp grid (3x3 first), repeat until resolution stops improving

Your resolution will likely stop improving somewhere during the above process, if not - lucky you! You have extremely strong data and are likely already at <4A or so and can test all the fancy things mentioned at the end of the tutorial :-)

Cheers,

Alister

--
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/1f5d6181-e066-4549-bcfb-44008864dc7an%40googlegroups.com.

[Paris Watson]

Hi Alister, 

Thank you this worked! doing the steps 1-3 I got 3.98, my nyquist is 3.95 so I think it is there. 

However I collected eer and was wondering if there would be any point to trying to upsample to 1/2 pixel size. Is that something thats possible in Warp/M? 

Thank you for your help! 

Paris 

[Alister Burt]

Congratulations! Great work :-)

You can do it but it might be a pain - have to go back and re-export motion corrected averages at the desired pixel size, 

 not sure exactly how much of your downstream processing would break… if you reconstruct tomograms at the same pixel spacing and everything lines up you should be good!

Only advised if the biology benefits from higher res

Cheers,

Alister

Sent from mobile - apologies for brevity

On Oct 28, 2025, at 09:06, Paris Watson <paris....@be-labs.org> wrote:

Hi Alister, 

To view this discussion visit https://groups.google.com/d/msgid/warp-em/75bb0006-fd6f-4ce0-b093-e185408c250bn%40googlegroups.com.