Biowulf High Performance Computing at the NIH
cryoDRGN: Deep Reconstructing Generative Networks for cryo-EM heterogeneous reconstruction

CryoDRGN is an algorithm that leverages the representation power of deep neural networks to directly reconstruct continuous distributions of 3D density maps and map per-particle heterogeneity of single-particle cryo-EM datasets. It contains interactive tools to visualize a dataset’s distribution of per-particle variability, generate density maps for exploratory analysis, extract particle subsets for use with other tools and generate trajectories to visualize molecular motions. CryoDRGN is open-source software freely available at


Important Notes

Interactive job
Interactive jobs should be used for debugging, graphics, or applications that cannot be run as batch jobs.

Allocate an interactive session and run the program. Sample session:

[user@biowulf]$ sinteractive --gres=gpu:k80:1
[user@cn4199 ~]$ module load cryoDRGN
+] Loading cuDNN/7.5/CUDA-10.0 libraries...
[+] Loading CUDA Toolkit  10.0.130  ...
[+] Loading singularity  3.8.5-1  on cn4199
[+] Loading cryoDRGN 0.3.4  ...
Copy test data to your current folder:
[user@user@cn4199 ~]$ cp -r $CRYODRGN_TEST/* .
Downsample test data:
[user@cn4199 ~]$ cryodrgn -h
usage: cryodrgn [-h] [--version]

CryoDRGN neural network reconstruction

optional arguments:
  -h, --help            show this help message and exit
  --version             show program's version number and exit

Choose a command:
[user@cn4199 ~]$ cryodrgn downsample -h
usage: cryodrgn downsample [-h] -D D -o MRCS [-b B] [--is-vol] [--chunk CHUNK]
                           [--relion31] [--datadir DATADIR]
                           [--max-threads MAX_THREADS]

Downsample an image stack or volume by clipping fourier frequencies

positional arguments:
  mrcs                  Input particles or volume (.mrc, .mrcs, .star, or

optional arguments:
  -h, --help            show this help message and exit
  -D D                  New box size in pixels, must be even
  -o MRCS               Output projection stack (.mrcs)
  -b B                  Batch size for processing images (default: 5000)
  --is-vol              Flag if input .mrc is a volume
  --chunk CHUNK         Chunksize (in # of images) to split particle stack
                        when saving
  --relion31            Flag for relion3.1 star format
  --datadir DATADIR     Optionally provide path to input .mrcs if loading from
                        a .star or .cs file
  --max-threads MAX_THREADS
                        Maximum number of CPU cores for parallelization
                        (default: 16)
[user@cn4199 ~]$ cryodrgn downsample data/toy_projections.mrcs -D 24 -o particles.24.mrcs
2021-02-24 10:09:00     Processing image 0 of 1000
2021-02-24 10:09:01     (1000, 24, 24)
2021-02-24 10:09:01     Saving particles.24.mrcs
Parse image poses from a consensus homogeneous reconstruction:
[user@cn4199 ~]$ cryodrgn parse_pose_star -h
usage: cryodrgn parse_pose_star [-h] -D D [--relion31] [--Apix APIX] -o PKL

Parse image poses from RELION .star file

positional arguments:
  input        RELION .star file

optional arguments:
  -h, --help   show this help message and exit
  -D D         Box size of reconstruction (pixels)
  --relion31   Flag for relion3.1 star format
  --Apix APIX  Pixel size (A); Required if translations are specified in
  -o PKL       Output pose.pkl
[user@cn4199 ~]$ cryodrgn parse_pose_star data/ -o test.pkl -D 300 --relion31 --Apix 1.03
2021-02-24 10:19:42     5 particles
2021-02-24 10:19:42     Euler angles (Rot, Tilt, Psi):
2021-02-24 10:19:42     [-102.30296    82.318041  124.706463]
2021-02-24 10:19:42     Converting to rotation matrix:
2021-02-24 10:19:42     [[ 0.81941806 -0.10080704  0.56426234]
 [-0.53288075  0.22868946  0.81470193]
 [-0.21116853 -0.96826601  0.13367414]]
2021-02-24 10:19:42     Translations (pixels):
2021-02-24 10:19:42     [ 0.01876117 -0.62801942]
2021-02-24 10:19:42     Writing test.pkl
Parse CTF parameters from a .star/.cs file:
[user@cn4199 ~]$ cryodrgn parse_ctf_star -h
usage: cryodrgn parse_ctf_star [-h] --Apix APIX -D D [--relion31] -o O
                               [--png PNG] [--kv KV] [--cs CS] [-w W]
                               [--ps PS]

Parse CTF parameters from a RELION .star file

positional arguments:
  star         Input

optional arguments:
  -h, --help   show this help message and exit
  --Apix APIX  Angstroms per pixel
  -D D         Image size in pixels
  --relion31   Flag for relion3.1 star format
  -o O         Output pkl of CTF parameters
  --png PNG    Optionally plot the CTF

Overwrite CTF parameters:
  --kv KV      Accelerating voltage (kV)
  --cs CS      Spherical abberation (mm)
  -w W         Amplitude contrast ratio
  --ps PS      Phase shift (deg)
[user@cn4199 ~]$ cryodrgn parse_ctf_star data/ -D 300 --Apix 1.03 -o ctf.pkl --relio
n31  --kv 10 --cs 1 -w 0.5
2021-02-24 10:24:39     5 particles
2021-02-24 10:24:39     Overriding accerlating voltage with 10.0 kV
2021-02-24 10:24:39     Overriding spherical abberation with 1.0 mm
2021-02-24 10:24:39     Overriding amplitude contrast ratio with 0.5
2021-02-24 10:24:39     CTF parameters for first particle:
2021-02-24 10:24:39     Image size (pix)  : 300
2021-02-24 10:24:39     A/pix             : 1.03
2021-02-24 10:24:39     DefocusU (A)      : 13108.082418
2021-02-24 10:24:39     DefocusV (A)      : 12845.582418
2021-02-24 10:24:39     Dfang (deg)       : -160.39
2021-02-24 10:24:39     voltage (kV)      : 10.0
2021-02-24 10:24:39     cs (mm)           : 1.0
2021-02-24 10:24:39     w                 : 0.5
2021-02-24 10:24:39     Phase shift (deg) : 0.0
2021-02-24 10:24:39     Saving ctf.pkl
[user@cn4199 ~]$  cryodrgn downsample data/ -D 24 -o relion31.24.mrcs --relion31
`fused_weight_gradient_mlp_cuda` module not found. gradient accumulation fusion with weight gradient computation disabled.
2022-03-15 18:02:03     Warning: /gpfs/gsfs7/users/user/cryoDRGN/relion31.24.mrcs already exists. Overwriting.
2022-03-15 18:02:03     Processing batch 0
2022-03-15 18:02:03     (5, 24, 24)
2022-03-15 18:02:03     Saving /gpfs/gsfs7/users/user/cryoDRGN/relion31.24.mrcs
[user@cn4199 ~]$ exit
salloc.exe: Relinquishing job allocation 59748321
[user@biowulf ~]$

Batch job
Most jobs should be run as batch jobs.

Create a batch input file (e.g. For example:

module load CryoDRGN      
cp -r $CRYODRGN_TEST/* .
cryodrgn parse_pose_star data/ -o test.pkl -D 300 --relion31 --Apix 1.03
cryodrgn parse_ctf_star data/ -D 300 --Apix 1.03 -o ctf.pkl --relio
n31  --kv 10 --cs 1 -w 0.5

Submit this job using the Slurm sbatch command.

sbatch [--cpus-per-task=#] [--mem=#]