Negative stain EM workflow

Home > Negative stain EM workflow

Negative stain EM workflow

Below is a typical workflow for a negative stain project that already has an initial model provided (through PDB/EMDB).

1. Collect negative stain dataset:

• Typically, a reasonable dataset will have 10,000 to 30,000 particles for negative stain EM analysis. You can estimate the number of micrographs needed by approximating how many 'good' particles are present in a given EM micrograph.

2. Picking particles

• Automatic picking using Signature:
  • In order to Signature, you will need to first have initial class averages to use as templates.
  • To get these templates, following these steps:
    • Manually pick ~1,000 particles using e2boxer.py, saving .box files for each micrograph picked.
    • Extract particles using makeStack.py (without CTF phase flipping)
    • Run Auto_Align.py on this initial stack, starting with ~100 class averages (--start) and finishing with 30 class averages (--final)
      • If class averages do not look very homogenous, trying picking more particles manually and repeating this process.
    • Select class averages to be used as templates for Signature, outputting these selected averages into a new stack
    • Use runSignature.py to pick automatically all of the particles in your dataset
• Alternatively, you can perform manual picking of your entire dataset, or semi-automated picking using SWARM in e2boxer.py.

3. Estimating CTF

• In order to perform CTF correction (phase flipping) you will need to first estimate the CTF of your micrographs using estimateCTF_CTFFIND3.py.
• This will generate a parameter file (or log files) containing per-micrograph CTF information.

4. Extracting particles with makeStack.py

• After picking particles & estimating the CTF, you will now need to extracting individual particles from the micrographs using makeStack.py
• Make sure to include the --phaseflip option when extracting particles, so that the program can use the CTF information per micrograph to phase the particles in the output stack.

5. 2D classification:

• With the CTF-corrected particles, you will now group the individual particles into class averages using Auto Align.py.
• This result will provide an unbiased snapshot of the structure(s) adopted by your sample.

6. 3D classification:

• Assuming that you have an initial 3D model to start your 3D analysis, you should try to perform 3D classification in Relion to identify a homogenous group of particles.
• To do this, run 3D classification in Relion withOUT CTF correction and group your data into 3 or 4 groups.
• If your data are extremely homogenous, then the models may all appear the same. However, it is more likely that one or two models will look higher resolution than the others. These particles that went into these models should be selected and used for 3D refinement.

7. 3D refinement:

• The next step involves further refining of the particles that were within the homogenous 3D classes.
• Select these particles using the Relion GUI and then perform 3D refinement.

8. Analysis

• In order to assess your model quality, you should compare projections of our 3D model with 2D class averages