Documentation for cryoDRGN2.

Installation

CryoDRGN2 commands are now natively available in the top-of-tree on github, and will be available in the next official cryoDRGN version 2.0 release.

• Older installation instructions

Running cryoDRGN2

There are two commands for ab initio reconstruction: cryodrgn abinit_homo (homogeneous ab initio reconstruction) and cryodrgn abinit_het (heterogeneous ab initio reconstruction):

# homogeneous ab initio
cryodrgn abinit_homo -h

# heterogeneous ab initio
cryodrgn abinit_het -h

Setup

• Downsample your particles to a box size of 128 either with cryodrgn downsample or with other tools.

• If you have a very large dataset (>500k images), I would recommend training on a subset of particles for initial testing. Use cryodrgn_utils select_random to select a random subset of particles.

  # get a random selection of 200k particles from a dataset of 1,423,124 particles
  cryodrgn_utils select_random 1423124 -n 200000 -o ind200k.pkl
  

  • You can then train on only the random subset with the argument --ind ind200k.pkl

• For reference, ab initio heterogeneous reconstruction of a recent dataset of ~218k 128x128 particles took 20 hours to train (1 V100 GPU).

Example usage

# homogeneous reconstruction
cryodrgn abinit_homo [particles] --ctf [ctf.pkl] -o [output_directory]  >> output.log

# heterogeneous reconstruction
cryodrgn abinit_het [particles] --ctf [ctf.pkl] --zdim 8 -o [output_directory]  >> output.log

Note on homogeneous reconstruction

We have found that cryoDRGN2’s homogeneous reconstruction can work better than traditional ab initio approaches for challenging heterogeneous datasets, e.g. the RAG1-RAG2 complex (EMPIAR-10049) and the pre-catalytic spliceosome (EMPIAR-10076), potentially due regularization properties of the neural model. It may be interesting to try cryodrgn abinit_homo for generating a consensus reconstruction or initial model. Please let us know if you find something interesting!

Note on training settings

• The default translational search extent is +/- 10 pixels. If your particles are not well-centered, you can use a wider search extent, e.g. +/- 40 pixels ( --t-extent 40).
• Poses are updated every 5 epochs (change this setting with --ps-freq)
• The default pose search settings are not tuned to produce super high-resolution volumes (a tradeoff of accuracy vs. compute speed).
• The default training time is 30 epochs. A typical use case is to run for 30 epochs, check the results (cryodrgn analyze), then extend training to 60 epochs. You can extend by rerunning with -n 60 --load latest. If your dataset is very large, you may want to reduce --ps-freq and -n.