Dear Chimera users. I am new to using chimera, so this question may seem a bit simplistic. I'm trying to use Chimera to examine EM tomographs of fairly busy data sets. I'd like to be able to divide the data into separate features so that I can better understand which parts are noise and which are structural (segmentation). So far I have found the "Pick Surface Pieces" tool is very useful for colouring the surface of features of interest. I have also found it useful to colour some surfaces with 0 opacity to remove items that are in the way, but I would prefer to do this in a more permanent fashion sot that I can return to a data set and quickly compare structures. I have also tried using "Volume Path Tracker" with "Color Zone" and "Split Map by Color" (after adjusting the voxel size to angstroms rather than nm to give a reasonable volume segmentation). My question is would it be possible to alter "Pick Surface Pieces" so that it colours (or tags in some way) all of the voxels within the chosen mesh, so that it would then be possible to split the map into separate compartments that could be saved and examined separately? This would also offer the ability to use different threshold levels to examine different features. I realise that this is more of a wish list than a trouble shooting question, but I'd also appreciate any input from more experienced Chimera/tomography users. Thank you for your help. Toby S -- Dr T Starborg FLS, University of Manchester Michael Smith Building, Rumford Street, Manchester,M13 9PL. UK tel:+44(0)161 275 1557 fax +44(0)161 275 1505
Hi Toby, Chimera does not have very useful tools for segmenting EM tomography although we would like to develop such tools. The pick surface pieces tool seems quite limited because it only works with connected contour surfaces. Just varying contour level does not give you much control. It would be possible to improve that tool as you suggested so that it copies the enclosed voxels to a new map, or sets the enclosed voxels to zero. I have some doubts that that would be useful. If you have to raise the contour level to break undesired connections, then you won't get a copy of the data out to low enough threshold levels around the pieces you choose. A couple tools that can help with segmentation are volume eraser, and zones in volume viewer. Volume eraser lets you set to zero map values within (or outside) a sphere that you can move around and resize. You can save the resulting map to a new file with volume dialog menu entry File / Save map as.... http://www.cgl.ucsf.edu/chimera/tutorials/volumetour/volumetour.html#handera... The volume dialog menu entry Features / Zone lets you show only parts of map within a radius of selected markers. Using volume dialog File / Save map as... will then save just the shown zone with zeros outside the given radius. http://www.cgl.ucsf.edu/chimera/tutorials/volumetour/volumetour.html#eospher... Sometimes you want zones around volume path tracer links, not just around markers. The undocumented bondzone command allows this. It is described below. The guide to volume data display describes other capabilities: http://www.cgl.ucsf.edu/chimera/tutorials/volumetour/volumetour.html Many labs use the commercial program Amira to segment EM tomography. You paint regions in individual data planes to delineate structures in the map. Tom Thomas Goddard wrote:
Hi Manfred,
I added the capability to segment out density around paths created with the Chimera path tracer tool. Formerly the zone capability only segmented density near marker positions, and not the connecting links. The new capability works by using closely spaced points placed along the links that join markers. Besides the volume dialog zone tool and equivalent surface zone tool (menu entry Tools / Surface & Binding Analysis / Surface Zone) the color zone tool (menu Tools / Volume Data / Color Zone) can now color the density surfaces near path tracer links.
This new feature is in Chimera version 1.2255. I did not figure out a good place to put the option in the dialogs and instead made a Chimera command, called "bondzone" to enable segmenting and coloring around links (also called "bonds"). The feature is currently not turned on by default. You type the command
bondzone
to the Chimera command entry field. To show the command entry field at the bottom of the main Chimera window use menu entry Favorites / Command Line. After typing this command the volume dialog zones, surface zone dialog, and color zone dialog all will use extra points (not displayed) along selected links. The spacing of the extra points is relative to the radius of the link cylinder. By default the points are spaced along the link separated by one link radius. You can change this using the bondzone command with a numeric argument. For example,
bondzone 0.5
will cause all subsequent segmenting and color zones to use points spaced at 0.5 times the radius of the link.
Note that only *selected* bonds and atoms are used by the zone tools. I added a new keyboard 2-key shortcut "sc" that extends the current selected atoms or bonds to all connected atoms and bonds. So you can click on one marker in a path tracer trace, then type sc to select all connected markers. Recall that to turn on keyboard shortcts you use menu entry Tools / General Controls / Accelerators On. They can be set on by default using the Prefereces dialog, Tools category. You can also press the up arrow key on the keyboard to extend a selection to all traces (connected or not) in the same path tracer "marker set".
If you want to turn the "bondzone" feature off so you get zones only around selected markers use command:
~bondzone
The tilde (~) in front of Chimera commands often means undo the normal command effect. When you turn bondzone on or off this is saved in your Chimera preferences file (~/.chimera/preferences) so that the setting remains the same when you next start Chimera.
Would be nice to not have to mess with this bondzone command and just have that be the default behaviour but it is currently pretty inefficient (ie slow) and I did not want to impair performance when taking a zone around a 10,000 atom crystal structure fit into a density map. Hopefully I can improve the performance in the future and make this zone behaviour the default.
Thanks for suggesting this capability.
Tom
Tom, Thank you for your quick reply. I have indeed looked at using Amira to segment our data, but currently we do not have the resources to purchase it with all the necessary modules. You have a good point about the possibility of removing too much data when trying to separate features by altering the threshold. My initial plan was to use the tools in a multi step process: The first step would be to use a crude threshold that selects the features of interest plus extra voxels. This step should hopefully eliminate many of the small particles that fill the tomograms (it could also be used to eliminate larger objects that you may not interested in, such as ribosomes when looking at intracellular structures). The idea was to use this crudely selected data set to find better thresholds with which to examine the features of interest. This should be made easier because there would be fewer small (1-2 voxel) particles crowding the view. A problem I can foresee is that the pick surface pieces currently surrounds the voxels on all surfaces, ie it can turn voxels surrounding a stick-like structures into tubes. To be able to use this method effectively the program would have to select the entire volume surrounded by the outer surface. I accept that this is probably quite a rough and ready way of segmenting the data, but it strikes me as a quick method of removing unconnected data from the features of interest that is less prone to human error than using the volume eraser. Thanks for the information on the bondzone. This feature seems useful. A quick extra question. Is it possible to increase the effective radius of colourzone? I have a data set with a voxel size of 1 nm cubed, so the current maximum radius of 30 angstroms only selects a radius of 6 voxels. I can reset the effective voxel size, but this makes the images look odd due to the large magnifications needed. Thanks again for all the help. TobyS -- Dr T Starborg FLS, University of Manchester Michael Smith Building, Rumford Street, Manchester,M13 9PL. UK tel:+44(0)161 275 1557 fax +44(0)161 275 1505
Hi Toby, You can use radii larger than 30 with the color zone tool but you have to type in the larger number. The slider only goes to 30 (was originally created for x-ray structures). Making the slider range adapt to the size of the displayed data is on my Chimera requests page (item # 60, http://www.cgl.ucsf.edu/chimera/plans.html). I'm still not convinced that selecting connected pieces of a density map is going to be useful. If you choose a low threshold to get the structure you want plus some extra padding, then it ends up being connected to all kinds of other structures you do not want. It will only work if your structure is not connected to other structures and that is rare. I'm interested in hearing segmentation tool ideas but reluctant to spend time on it if it will not be useful. I'll be on vacation tomorrow until June 27. Other Chimera developers can answer your questions on the chimera-users list although I am the one who develops the volume features. Tom
participants (2)
-
Toby Starborg -
Tom Goddard