Classification tutorial (EMPIAR-10304)
Contents
Classification tutorial (EMPIAR-10304)¶
This tutorial shows how to convert raw tilt-series from EMPIAR-10304 (E. coli. ribosomes) into a ~4.9A resolution structure and resolve 8 different conformations.
Total running time required to complete this tutorial: 32 hrs.
We first use the command line to download and decompress a tbz file containing: 1) a script to download the raw tilt-series from EMPIAR, 2) corresponding metadata with tilt angles and acquisition order, and 3) an initial model:
# cd to a location in the shared file system and run:
wget https://nextpyp.app/files/data/nextpyp_class_tutorial.tbz
tar xfz nextpyp_class_tutorial.tbz
source download_10304.sh
Note
Downloading the raw data from EMPIAR can take several minutes.
Open your browser and navigate to the url of your nextPYP
instance (e.g., https://nextpyp.myorganization.org
).
Step 1: Create a new project¶
Data processing runs are organized into projects. We will create a new project for this tutorial
The first time you login into
nextPYP
, you should see an empty Dashboard:Click on Create new project, give the project a name, and select Create
Select the new project from the Dashboard and click Open
The newly created project will be empty and a Jobs panel will appear on the right
Step 2: Import raw tilt-series¶
Import the raw tilt-series downloaded above ( <1 min)
Go to Import Data and select Tomography (from Raw Data)
A form to enter parameters will appear:
Go to the Raw data tab:
Set the
Location
of the raw data by clicking on the icon and browsing to the directory where the you downloaded the raw movie framesType
*.mrc
in the filter box (lower right) and click on the icon to verify your selection. 12 matches should be displayedClick Choose File Pattern to save your selection
Click on the Microscope parameters tab
Set
Pixel size (A)
to 2.1Set
Acceleration voltage (kV)
to 300Set
Tilt-axis angle (degrees)
to 90.0
Click Save and the new block will appear on the project page
The block is in the modified state (indicated by the sign) and is ready to be executed
Clicking the button Run will show another dialog where you can select which blocks to run:
Since there is only one block available, simply click on Start Run for 1 block. This will launch a process that reads one tilt image, applies the gain reference (if applicable) and displays the resulting image inside the block
Click inside the block to see a larger version of the image
Step 3: Pre-processing¶
Movie frame alignment, CTF estimation and particle picking ( 4 min)
Click on Tilt-series (output of the Tomography (from Raw Data) block) and select Pre-processing
Go to the Frame alignment tab:
Check
Single-file tilt-series
Click on the CTF determination tab
Set
Max resolution
to 5.0Click on the Tomogram reconstruction tab
Check
Erase fiducials
Set
Binning factor for reconstruction
to 12Set
Thickness of reconstruction (unbinned voxels)
to 3072Uncheck
Resize squares to closest multiple of 512
Click on the Resources tab
Set
Threads per task
to 42Set
Memory per task
to 100Set other runtime parameters as needed (see Computing resources)
Click Save, Run, and Start Run for 1 block. Follow the status of the run in the Jobs panel
Step 4: Particle detection¶
Particle detection from virion surfaces ( 2 min)
In the Pre-processing block, click on the menu icon and select the Edit option.
Go to the Particle detection tab:
Set
Detection radius (A)
to 80Set
Detection method
to autoSet
Threshold for contamination detection
to 2.0Set
Minimum contamination size (voxels)
to 60Set
Minimum distance between particles
to 2Check
Local refinement
Set
Z-axis detection range (binned voxels)
to 40Set
Particle detection threshold
to 2
Click Save, Run, and Start Run for 1 block
Navigate to the Reconstruction group inside the Tilt-series tab to inspect the picked coordinates
Step 5: Reference-based refinement¶
Reference-based particle alignment ( 26 hr)
Click on Particles (output of the Pre-processing block) and select Particle refinement
Go to the Sample tab:
Set
Molecular weight (kDa)
to 2000Set
Particle radius (A)
to 150Click on the Extraction tab
Set
Box size (pixels)
to 64Set
Image binning
to 4Click on the Refinement tab
Specify the location of the
Initial model
by clicking on the icon , navigating to the folder where you downloaded the data for the tutorial, and selecting the file EMPIAR-10304_init_ref.mrcClick in
Alignments from sub-volume averaging
to select the initial parameters text filetomo-preprocessing-*_original_volumes.txt
from Pre-processingSet
Max resolution (A)
to 22.0Check
Use signed correlation
Check
Skip refinement
Click on the Constrained refinement tab
Set
First exposure for refinement
to 15Set
Last exposure for refinement
to 25Set
Optimizer - Exhaustive search points
to 5000000Check
Refine particle alignments
Set
Phi range (degrees)
,Psi range (degrees)
andTheta range (degrees)
to 180Set
Translation range (voxels)
to 50Check
Invert CTF handedness
Click on the Reconstruction tab
Set
Max tilt-angle
to 50Set
Min tilt-angle
to -50Click on the Resources tab
Set
Threads per task
to the maximum allowable by your systemSet
Memory per task
to at least 4x the number ofThreads per task
Set
Walltime per task
to 72:00:00
Save your changes, click Run and Start Run for 1 block
One round of refinement and reconstruction will be executed. Click inside the block to see the results
Step 6. Filter particles¶
Identify duplicates and particles with low alignment scores ( 3 min)
Click on Particles (output of the Particle refinement block) and select Filter particles
Go to the Particle filtering tab:
Set
Score threshold
to 15.0Set
Min distance between particles (A)
to 20Specify the location of
Input parameter file
by clicking on the icon and selecting the file tomo-coarse-refinement-*_r01_02.par.bz2Set
Lowest tilt-angle
to -7.0Set
Highest tilt-angle
to 7.0Set
Min number of projections per particle
to 1Check
Generate reconstruction after filtering
Click on the Refinement tab
Specify the location of the
Initial model
by clicking on the icon and selecting the file tomo-coarse-refinement-*_r01_02.mrc
Click Save, Run, and Start Run for 1 block. You can see how many particles were left after filtering by looking at the job logs.
Step 7 (optional): Permanently remove bad particles¶
Permanently remove bad particles to improve processing efficiency downstream ( 1 min)
Edit the settings of the existing Filter particles block
Go to the Particle refinement tab:
Check
Permanently remove particles
Uncheck
Generate reconstruction after filtering
Click Save, Run, and Start Run for 1 block to launch the job
Step 8. Fully constrained refinement¶
Tilt-geometry parameters and particle poses are refined in this step ( 10 min)
Click on Particles (output of the Particle filtering block) and select Particle refinement
Set
Box size (pixels)
to 256Set
Image binning
to 1Click on the Refinement tab
Specify the location of the
Initial model
by clicking on the icon and selecting the file tomo-fine-refinement-*_r01_02.mrcSelect the location of the
Initial parameter file
by clicking on the icon and selecting the file tomo-fine-refinement-*_r01_02.par.bz2 (select the filetomo-fine-refinement-*_r01_02_clean.par.bz2
if bad particles were permanently removed in the previous step)Set
Max resolution (A)
to 18:14Check
Use signed correlation
Set
Last iteration
to 3Check
Skip refinement
Click on the Constrained refinement tab
Set
First exposure for refinement
to 15Set
Last exposure for refinement
to 25Set
Optimizer - Exhaustive search points
to 0Set
Optimizer - Max step length
to 100Check
Refine tilt-geometry
Check
Refine particle alignments
Set
Phi range
,Psi range
andTheta range
to 30.0Set
Translation range (voxels)
to 30.0Click on the Exposure weighting tab
Check
Dose weighting
Check
Global weights
Set
Frame weight fraction
to 4
Click Save, Run, and Start Run for 1 block to execute three rounds of refinement and reconstruction
Click inside the Particle refinement block to inspect the results
Step 9: Create shape mask¶
Use most recent reconstruction to create a shape mask ( <1 min)
Click on Particles (output of Particle refinement block) and select Masking
Go to the Masking tab:
Select the
Input map
by click on the icon and selecting the file tomo-coarse-refinement-*_r01_03.mrcSet
Threshold for binarization
to 0.4Check
Use normalized threshold
Set
Width of cosine edge (pixels)
to 8
Click Save, Run, and Start Run for 1 block to run the job
Click on the menu icon of the Masking block, select the Show Filesystem Location option, and Copy the location of the block in the filesystem (we will use this in the next step))
Click inside the Masking block to inspect the results of masking
Step 10. Region-based local refinement¶
Constraints of the tilt-geometry are applied over local regions ( 25 min)
Edit the settings of the existing Particle refinement block and go to the Refinement tab:
Go to the Refinement tab:
Set
Max resolution (A)
to 18:14:12:10:8:6:5Set
Last iteration
to 8Specify the location of the
Shape mask
produced in Step 10 by clicking on the icon , navigating to the location of the Masking block by copying the path we saved above, and selecting the file frealign/maps/mask.mrcClick on the Constrained refinement tab
Set
First exposure for refinement
to 18Set
Last exposure for refinement
to 22Set
Number of regions
to 8,8,2Set
Tilt-angle range (degrees)
andTilt-axis range (degrees)
to 5.0Set
Phi range
,Psi range
andTheta range
to 5.0Set
Translation range (voxels)
to 20.0
Click Save, Run, and Start Run for 1 block to run the job
Click inside the Particle refinement block to inspect the results
Step 11: Particle-based CTF refinement¶
Per-particle CTF refinement using most recent reconstruction ( 2 hr)
Edit the settings of the existing Particle refinement block and go to the Refinement tab:
Go to the Refinement tab:
Set
Max resolution (A)
to 18:14:12:10:8:6:5:4.5Set
Last iteration
to 9Click on the Constrained refinement tab
Set
First exposure for refinement
to 15Set
Last exposure for refinement
to 25Uncheck
Refine tilt-geometry
Uncheck
Refine particle alignments
Check
Refine CTF per-particle
Set
Defocus 1 range (A)
andDefocus 2 range (A)
to 2000.0
Click Save, Run, and Start Run for 1 block
Click inside the Particle refinement block to inspect the results
Step 12: Region-based refinement after CTF refinement¶
Constraints of the tilt-geometry are applied over local regions ( 20 min)
Edit the settings of the existing Particle refinement block and go to the Refinement tab:
Set
Max resolution (A)
to 18:14:12:10:8:6:5:4.5:6:5:4.5Set
Last iteration
to 12Click on the Constrained refinement tab
Set
First exposure for refinement
to 18Set
Last exposure for refinement
to 22Set
Number of regions
to 16,16,4Set
Optimizer - Max step length
to 20.0Check
Refine tilt-geometry
Set
Translation range (pixels)
to 20.0Check
Refine particle alignments
Uncheck
Refine CTF per-particle
Click on the Exposure weighting tab
Set
Frame weight fraction
to 2
Click Save, Run, and Start Run for 1 block to run the job
Click inside the Particle refinement block to inspect the results
Step 13: 3D classification¶
Constrained classification ( 3 hr)
Click on Particles (output of the Particle refinement block) and select Particle refinement to create a new block
Go to the Refinement tab:
Specify the location of the
Initial model
by clicking on the icon and selecting the file tomo-coarse-refinement-*_r01_12.mrcSelect the location of the
Initial parameter file
by clicking on the icon and selecting the file tomo-coarse-refinement-*_r01_12.par.bz2Set
Max resolution (A)
to 8Set
Last iteration
to 20Click on the Constrained refinement tab
Uncheck
Refine tilt-geometry
Uncheck
Refine particle alignments
Click on the Classification tab
Set
Number of classes
to 8Click on the Exposure weighting tab
Specify the location of the
External weights
by clicking on the icon and selecting the file frealign/weights/global_weights.txt from the file location of the previous block
Click Save, Run, and Start Run for 1 block
Click inside the Particle refinement block to inspect the results
Tip
Click on the round blue markers (top right of the page) to inspect different classes or go to the Class view or Classes Movie tabs to show all classes simultaneously
Note
Running times were measured running all tilt-series in parallel on nodes with 124 vCPUs, 720GB RAM, and 3TB of local SSDs