Template picking with downscaled templates

Hi all,
When I do a first round of 2D classification and ab initio, but with downscaled particles, can I still use the map, create templates and use them for template picking?


Yes no problem with that (you can also just use the 2D classes rather than projections of the ab initio model)

1 Like

How does it work behind the scenes then? Are templates upscaled to match the micrographs?

you can also just use the 2D classes rather than projections of the ab initio model)

Thing is, its super difficult to see what are actually views from the protein just from the 2D classes alone. Therefore I tried to take the route via ab inito 10 classes simply on all particles (downscaled from 352 px to 48 px). One class did show the protein. I did NU refinment and now want to use the templates created from this map for a new round of picking

The templates are already at the same scale (in Å) as the micrographs, just with different binning.

They are lowpass filtered prior to searching, I am not sure if any other scaling or normalization is performed.

If one 3D class shows the protein, then a 2D classification on that subset of particles should also yield suitable templates.

Can you clearly see your particles on the micrograph?

How are the already at the correct scale? I downscaled them by a factor of 5. If they’re not upscaled back to it’s original scale, then the templates are 5 times smaller then the actual particles in the micrograph.
What am I missing?

Particles I can see clearly in raw micrographs, as it is phase plate data, I even see the correct shape

Downscaling does not change the size of the particles (in Å). It changes the number of pixels. For example, imagine I have a 200Å box at bin 1, 1Å/pixel - so a 200px box. I then downscale by 5. The box is now 40 pixels, but at a scale of 5Å/pixel - the box size is still 200Å. Make sense?