|
Size: 2271
Comment:
|
Size: 2283
Comment:
|
| Deletions are marked like this. | Additions are marked like this. |
| Line 5: | Line 5: |
| Trial and error. This is explained a bit in the proc3d manual page, but here it is again rewritten. amask= is very much like a flood-fill (the paint-bucket icon) in most 2-D paint programs. The idea is you start with some point, then fill in anything touching that point above some threshold value. Since it's annoying to specify a specific point in 3-D, and the maps are centered, we start in the middle of the map. However, sometimes maps are hollow in the middle, so the first parameter is the radius of a sphere centered on the origin that's big enough to touch some part of your structure. It should definitely NOT be so large that it extends outside your structure at any point. It just needs to be large enough to touch your structure somewhere. |
Automasking performs a procedure very much like solvent flattening, and can be critical for obtaining an optimal reconstruction. The only case where it should be avoided is where there are flexible or low density structures sticking out of your map which you want to insure you don't accidentally remove. |
| Line 8: | Line 8: |
| The second parameter is an isosurface threshold for flood filling. ie - if you display your map in Chimera or some other isosurface visualization program, this is the threshold you use to display your map so it looks nice. Typically, since you want to have some flexibility in how you display the map, you will actually want to use a somewhat lower value than this. ie - the lowest value you would possibly use when presenting the map. So long as any noise outside the structure isn't connected to the structure itself at the threshold you use, you're fine. |
Here is the procedure for determining automask parameters. There is also some information in the proc3d manual page under automask2= : |
| Line 12: | Line 10: |
| Since we don't want to put a sharp edged mask on our structure at an isosurface value this high, the 3rd parameter allows us to extend the mask we've constructed by building 1-voxel wide shells outside the existing mask. If you say 10 for the final number, it will add 10 shells to the mask. An important point is that in these added shells decay gradually (a Gaussian) to the background value. This prevents the masking from exaggerating the resolution of your map when used in an even/odd test. Typically the 3rd value should be roughly the box size/10 (in pixels). | Visualize your map in Chimera or some other isosurface visualization program. Pick an isosurface threshold at which your map 'looks good', then lower the threshold a little below that point. You want to avoid lowering it so much that any blobs of noise outside the map become connected to the bulk of the map itself. However, you also need to insure that the density representing your protein is fully interconnected. Typically 0.1-0.3 below the 'good' threshold works well. This isosurface threshold is the SECOND parameter for amask=. The first parameter for amask= is a radius in pixels. This should specify the size of a sphere which JUST BARELY comes into contact with the inside edge of your model. This can be done by trial and error. In many cases 2-3 pixels will be fine, but if your map is hollow, you will need a larger value. The third parameter specifies the width in pixels of a soft mask shell which goes around the outside of your particle. It helps insure that masking doesn't produce any resolution exaggeration. Typically this number is box_size_in_pixels/10, but you can go +-50% from this value. To test your 3 parameters: proc3d my_map.mrc test_map.mrc automask2=<r>,<threshold>,<width> then: v4 my_map.mrc test_map.mrc this will let you visualize the map and satisfy yourself that you aren't removing anything inappropriate. When you have the parameters you like, in refine, add: amask=<r>,<threshold>,<width> |
Q: How do I choose the correct arguments for the amask argument in refine? I see that it ultimately is used to run the amask2 argument in proc3d, but I'm still a little unsure as do how to choose the parameters. I take it that the 'r' is the approximate radius where the density of the molecule falls off, and thr is the value of ...? And what is N? Thank you.
A: Automasking performs a procedure very much like solvent flattening, and can be critical for obtaining an optimal reconstruction. The only case where it should be avoided is where there are flexible or low density structures sticking out of your map which you want to insure you don't accidentally remove.
Here is the procedure for determining automask parameters. There is also some information in the proc3d manual page under automask2= :
Visualize your map in Chimera or some other isosurface visualization program. Pick an isosurface threshold at which your map 'looks good', then lower the threshold a little below that point. You want to avoid lowering it so much that any blobs of noise outside the map become connected to the bulk of the map itself. However, you also need to insure that the density representing your protein is fully interconnected. Typically 0.1-0.3 below the 'good' threshold works well. This isosurface threshold is the SECOND parameter for amask=.
The first parameter for amask= is a radius in pixels. This should specify the size of a sphere which JUST BARELY comes into contact with the inside edge of your model. This can be done by trial and error. In many cases 2-3 pixels will be fine, but if your map is hollow, you will need a larger value.
The third parameter specifies the width in pixels of a soft mask shell which goes around the outside of your particle. It helps insure that masking doesn't produce any resolution exaggeration. Typically this number is box_size_in_pixels/10, but you can go +-50% from this value.
To test your 3 parameters:
proc3d my_map.mrc test_map.mrc automask2=<r>,<threshold>,<width>
then: v4 my_map.mrc test_map.mrc
this will let you visualize the map and satisfy yourself that you aren't removing anything inappropriate. When you have the parameters you like, in refine, add: amask=<r>,<threshold>,<width>