Skip to content

Latest commit

 

History

History
116 lines (86 loc) · 6.81 KB

Doc.md

File metadata and controls

116 lines (86 loc) · 6.81 KB
Raw

Doc

Open the structure and the volume map, open the Interactive tab, click Fit.

The whole UI design is with the hope to allow users to be able to use DiffFit without training. This is especially the case if the user is familiar with ChimeraX's Fit in Map tool or fitmap command. We provide a detailed Doc below for different usage scenarios.

Heads-up

We suggest using DiffFit as an initial fitting tool and then using the ChimeraX built-in "Fit in Map" to refine DiffFit's results. Because we observed in most cases, the "Fit in Map" refined results' RMSD is slightly better than DiffFit's.

A CUDA-compatible GPU will give the best performance. But you can also run DiffFit on the CPU. If you run on the CPU, you may click Fast in the Options button to get the results faster. You may even decrease # shifts and/or # quaternions if you are really in a hurry. However, the chance that you find a plausible fit is lower.

Demo usage scenarios

Scenario 1: Fit a single structure

YouTube Tutorial

  1. Open the source structure and the target map in ChimeraX. Let's use PDB-6WTI and its associated EM Map EMD-21897 as an example.
    1. Option 1, via ChimeraX command line. Run the following command in the command line at the very bottom of the ChimeraX windows.
      1. open 6WTI
      2. open 21897 from emdb
    2. Option 2, via download the files and drag and drop.
      1. You can find the official RCSB webpage for a PDB ID by either Googling the ID or composing a URL: https://www.rcsb.org/structure/6WTI. Change 6WTI to another ID if needed.
      2. Download Files > PDBx/mmCIF Format
      3. Click the first EMDB, after EM Map EMD-ID to access its EM map webpage
      4. Download > 3D volume (map.gz); unzip the downloaded map file.
      5. Drag and drop both files (6wti.cif and emd_21897.map) into the ChimeraX window.
  2. Change the iso-surface threshold level of the map.
    1. The Volume Viewer is usually located at the bottom-right of the ChimeraX window. Move the slide to change.
    2. DiffFit is very robust against this parameter, so very often, you don't have to change it.
    3. To get the best performance, change the level to a value where you can see some secondary structures (alpha-helices or beta sheets).
  3. In the DiffFit panel, go to the Interactive tab. Click Fit.
  4. After computing, DiffFit will automatically go to the View tab and select the top fit. You may change the threshold values. Usually, the defaults work fine.
  5. You can now click on the rows to go through the fitting results. You may sort the table by a different metric by clicking the header (by default, the table is sorted by Density).
  6. Once you find a plausible fit, you may use ChimeraX's Fit in Map tool (or a command similar to fit #1 in #2) to refine the placement.
  7. If you have a ground truth structure to compare with, you may open that structure and use a command similar to rmsd #1 to #3 to calculate the RMSD. Check the RMSD doc for more.
  8. Save a molecule by clicking Structure if desired; and zero the density occupied by the current molecule:
    1. Click Simulate volume
    2. Change the surface level threshold for the simulated volume if necessary
    3. Click Zero density
  9. Repeat the last step (Save, Simulate, Zero) for other plausible fits
  10. Save the last working volume by File > Save > Files of type: MRC > Map: working volume and use it as the input for another round if needed.
  11. In the Interactive tab, you may click Options to change the fitting parameters.
  12. In the Settings tab, you may change the global settings of DiffFit. The benchmark table in our paper uses Fit atoms: All atoms on an Nvidia RTX 4090 GPU.
  13. You can find the computing time in the ChimeraX log window, usually at the right. Look for DiffFit total time elapsed: . The first run is slightly slower as there are some global initialization processes.
  14. If you are interested in comparing DiffFit with the ChimeraX Fit in Map command, you may run a command similar to fit #1 in #2 search 1000. Check the fitmap doc for more.

Scenario 2: Composite multiple structures

YouTube Tutorial

The logic of compositing is you fit multiple structures in one run to one Cryo-EM volume map. Apart from the target volume file, you will also need to prepare all the individual structures that you want to fit and simulate a map for each structure. You will run this functionality in Disk mode by specify the path to the involved files or folders.

As a demo, let's composite the individual chains from PDB-8SMK into EMD-40589.

  1. Open 8SMK in ChimeraX

  2. Go to the Utilities tab. Under the Split a structure into individual chains section:

    1. Set an Output Folder. The default is a new folder called split_out in your working directory, which usually is your desktop.
    2. Select 8smk.cif as the Structure.
    3. Press Split.
    4. Optional: Go to the output folder, delete chain D, E, and F. Because they are the same as chain A, B, and C.

  3. Go to the Utilities tab. Under the Simulate a map for each structure in the folder section:

    1. If your individual chains are in split_out. Then you may directly click Simulate. By default, the simulated maps are in a new folder called sim_out in your working directory, which usually is your desktop.

  4. Go to the Disk tab, set all the file and folder paths, set the Target Surface Threshold (you may open the map in the same ChimeraX window to decide a good surface level value).

  5. Click Run!.

  6. Use the same way as in Scenario 1: Fit a single structure to view the results.

  7. Save the working volume and go for another round of fitting if needed.

Scenario 3: Identify unknown densities

YouTube Tutorial

The whole procedure is similar to Scenario 2: Composite multiple structures, only that the input files are different.

A quick demo search from 3 structures

All the necessary files are at: https://github.com/nanovis/DiffFit/tree/main/dev_data/input/domain_fit_demo_3domains.

Search from the whole demo library

Use the volume map from the above link.

Use the structure files from https://osf.io/download/vgk7f/.

The computing time for searching the whole candidate library on one RTX 4090 is about 7 minutes.