EMAN Wiki

User Tools

Site Tools

Trace:
eman2:e2tomosmall

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision		Previous revision
eman2:e2tomosmall [2025/08/05 01:08] – [Tomogram annotation (GPU recommended)] steveludtkeeman2:e2tomosmall [2025/08/05 11:06] (current) – [Particle extraction (~2 min (a few manual) - hours (a couple of thousand))] steveludtke
Line 198: Line 198:
 This is the easiest approach by far, and is basically a single step process, but is only available in recent (2022+) snapshots of EMAN2. This is the easiest approach by far, and is basically a single step process, but is only available in recent (2022+) snapshots of EMAN2.
   * You should easily be able to select ~500 particles per tomogram, potentially more, but 500 is more than sufficient for the tutorial   * You should easily be able to select ~500 particles per tomogram, potentially more, but 500 is more than sufficient for the tutorial
-  * For instructions, see: [[EMAN2:e2tomo_more|Automated Particle Selection]] Section 
-  * When complete, skip ahead to the next major section **Particle Extraction** 
  
 +{{https://blake.bcm.edu/dl/EMAN2/sptboxer_convnet.png}}
 +
 +  - Subtomogram Averaging → Convnet based auto-boxing
 +    * label: **ribo**  (this can be anything you like, but you will need to use the same label later)
 +    * gpuid: **-1** //if you don't have a GPU, otherwise use default//
 +  - **Launch**
 +  - Four windows will appear: "tomobox.py", "(Negative)", "(Positive)", "(Particles)". Arrange on the screen so you can reasonably access them. One more window will appear in a moment.
 +  - The "tomobox.py" window contains controls on the left and a summary of the available tomograms on the right. 
 +  - Click on tomob. This will cause a window showing the central section of the tomogram to appear. You will need this a lot, so move it somewhere where it is accessible, but doesn't cover everything else.
 +  - Next to the "New" button, select "Good References" from the menu. With this selected, clicking on the tomogram will select positive references.
 +  - The goal is to select a few (~5 - 20) regions with a ribosome particle well centered in the middle of the circle (which will appear when you click). 
 +    * Each time you click, the corresponding region will appear in the (Positive) window
 +    * Holding shift and clicking will remove a reference already selected.
 +    * Using the up/down arrows in the tomogram view will allow you to move up and down through the Z slices of the tomogram.
 +  - Once you have selected 5 or more well centered particles, Change "Good References" to "Bad References".
 +  - Use the arrow keys to move "above" or "below" the layer where particles appear. 
 +    * Do **not** select very high contrast "bad" particles, like fiducials. 
 +    * Instead, select regions of background which contain nothing at all, or weak features which don't look like a particle. 
 +    * Again, select 5 or more of these background regions, which will appear in the (Negative) window.
 +  - Once you're happy with your references, hit the **Train** button.
 +    - Bring the terminal window where you launched the project manager to the front. As it trains you will see a series of lines like: "iteration 0, cost 1.715". In each iteration, you hope for the cost to decrease. If it instead starts at ~2.0 and doesn't change much, when the training completes (Niter) try to select better references and train again.
 +  - Assuming the number decreases to a reasonable value (<1.0). You probably got a pretty decent result. 
 +  - Change "Bad References" to "Particles". Try pressing **Apply**.
 +  - You should see some particles appear in both the tomogram (you may need to move up and down) and the (particles) window. The table will tell you how many were found. 
 +  - If the particles look ok, but the number found is significantly less than the ~500 you should be able to get, try reducing "PtclThres" in increments of 0.1, pressing "Apply" after each change. The number should increase. Don't go much above 500.
 +  - If you are satisfied with the results on this Tomogram, press **Save** to save the network, then press **ApplyAll** to box out all 3 Tomograms.
 +  - When complete, skip ahead to the next major section **Particle Extraction**
 +
 +<note>
 +See: [[EMAN2:e2tomo_more|Automated Particle Selection]] Section for the original instructions for this tool, but they may be a bit dated.
 +</note>
 ==== Tomogram annotation (GPU recommended) ==== ==== Tomogram annotation (GPU recommended) ====
 This is an older strategy using the deep-learning based tomogram annotation program to find particles. It still works, but the new deep-learning picker in the section above will generally work better. This is an older strategy using the deep-learning based tomogram annotation program to find particles. It still works, but the new deep-learning picker in the section above will generally work better.
Line 255: Line 284:
   * If you did the previous optional annotation step above, you will be able to see the selected particles here, and if you like, manually update them.   * If you did the previous optional annotation step above, you will be able to see the selected particles here, and if you like, manually update them.
  
-===== Particle extraction (~2 min (a few manual) - hours (a couple of thousand)) =====+===== Particle extraction (~2 min (a few manual) - over an hour (a couple of thousand, large tutorial)) =====
 Note that this step will be vastly different resource-wise if you are only extracting a few manually selected particles for purposes of later template matching or if you already selected hundreds of particles from each tomogram.  Note that this step will be vastly different resource-wise if you are only extracting a few manually selected particles for purposes of later template matching or if you already selected hundreds of particles from each tomogram. 
 +
 +<note>
 +Depending on how you selected particles you may have a significant number of particles which contain fiducials or artifacts from the fiducials (bright spots). Having these in your particle data can cause significant problems, and could lead to getting bad initial models and thus bad refinements. For a small data set like this, it might be a good idea to open the manual boxer for each tomogram and delete any particles with fiducials in them before completing this step (particle extraction).
 +</note>
  
 The reduced 1k x 1k (or 2k) tomograms are used only as a reference to identify the location of the objects to be averaged. Now that we have particle locations, the software returns to the original tilt-series, extracts a per-particle tilt-series, and reconstructs each particle in 3-D independently at full resolution. Since this is performing a full resolution reconstruction of each particle it is somewhat resource intensive. The reduced 1k x 1k (or 2k) tomograms are used only as a reference to identify the location of the objects to be averaged. Now that we have particle locations, the software returns to the original tilt-series, extracts a per-particle tilt-series, and reconstructs each particle in 3-D independently at full resolution. Since this is performing a full resolution reconstruction of each particle it is somewhat resource intensive.
Line 266: Line 299:
     * set //boxsz_unbin// to 128 (Small) or 256 (Large).     * set //boxsz_unbin// to 128 (Small) or 256 (Large).
       * If you had the correct size in the previous step this should be the same as leaving the default -1       * If you had the correct size in the previous step this should be the same as leaving the default -1
-    * enter the label you used when picking particles ("initribo" if you manually boxed, "tomobox" if you used the deep learning picker)+    * enter the label you used when picking particles ("initribo" if you manually boxed, "ribo" or "tomobox" if you used the deep learning picker)
     * //threads// = value for your machine     * //threads// = value for your machine
     * Launch     * Launch
Line 317: Line 350:
  
 Once it gets past 3-4 iterations, you can use the browser to look in //sptsgd_00//, and double-click on //output_cls0.hdf//. This file will change after each iteration completes. It contains the results of the most recent iteration. When you are satisfied with the quality of the initial model, you can kill it with the task manager in e2projectmanager. Once it gets past 3-4 iterations, you can use the browser to look in //sptsgd_00//, and double-click on //output_cls0.hdf//. This file will change after each iteration completes. It contains the results of the most recent iteration. When you are satisfied with the quality of the initial model, you can kill it with the task manager in e2projectmanager.
 +
 +Note that while the program converges pretty quickly, it won't always get the correct structure, particularly if you have some bad particles (like fiducials) included in the particle set. It is important to look at the resulting starting map and make sure it looks at least vaguely like you expect. If not you may wish to try running this step again. This will produce //sptsgd_01//, //02//, ...
  
 <note tip> <note tip>
eman2/e2tomosmall.1754356080.txt.gz · Last modified: by steveludtke