Masking in Frealign (using focus_mask )

Submitted by Axel on Wed, 2017-09-20 19:52

Forums

Frealign

I was wondering how the "focus_mask" parameter worked, and whether you had any advice on using this in refinement/classification, and how small a region I should expect to be able to work with.

Thanks a lot!

Axel

niko Wed, 2017-09-20 21:23

The focused mask is described in Grigorieff (2016). You have to define a sphere by giving an x,y,z coordinate (measured in Angstrom from the first voxel in the volume) and a radius (also in Angstrom). This is then projected along the direction of the view of the particle to define a circular region that is used to evaluate differences between the reference/class average and the particle. In other words, the focused mask excludes regions that do not contain relevant signal for the classification. A radius of 30 to 40 Angstrom should be used but you can also try smaller values and see if they give good results.

dlyumkis Thu, 2017-09-21 18:04

Axel, we do this all the time, just put a sphere (in Chimera) onto the area that you're interested in within your reconstruction. Then resample the sphere onto the grid of your reconstruction. Once the sphere is in the area you want and resampled, type:

measure center #1 mark true

where "#1" refers to the model number of the sphere, assuming your reconstruction is #0

That will give you the coordinates. I believe those values are in pixels (????, please double check). But in any case, as Niko mentioned, Chimera is using the same coordinate system, and the values (or scaled to the pixel size) can be directly used in focused classification.

himesb Thu, 2017-09-21 19:58

Although there may be an exception, I'm fairly certain chimera always reports angstrom. This of course depends on the pixel size in the header being correct.

Axel Fri, 2017-09-29 20:05

This is pretty much what I've been doing.

Glad to hear the coordinate systems are the same, and that this should work.

I'm submitting a new attempt at this today.

Axel