Scipion

Scipion is a workflow-based image processing framework for obtaining 3D models of macromolecular complexes from cryo-EM data, backed by remote compute. It integrates packages like RELION, Xmipp, EMAN2, and CTFfind behind a unified GUI, streamed here in a remote desktop session on the Ubuntu Desktop image.

Usage

1. Deploy

dxflow workflow create --identity scipion hub://scipion

# Start with defaults, or tune per run with --override
dxflow workflow start scipion
dxflow workflow start scipion \
    --override env.app.VNC_PASSWORD=my-strong-pass

# Publish the web port on an HTTPS link
dxflow workflow start scipion --link

2. Open the app

Open your browser at http://localhost:6082/vnc.html and enter the password you set in VNC_PASSWORD. The Scipion project manager opens on the desktop. Port 5901 is also exposed for connecting a native VNC client. A start given --link publishes port 6082 at an HTTPS URL printed on the start line — open it at /vnc.html to reach Scipion from anywhere.

3. Persist data

Mount your movies and datasets under /volume — the Scipion user data directory is linked there, so projects and results survive restarts.

Configuration

name: scipion
tags:
    - structural
steps:
    - name: app
      runtime: docker
      mode: parallel
      image: ghcr.io/dxflow-ai/scipion:latest
      volumes:
          - name: volume
            host: ./volume
            container: /volume
      ports:
          - name: web
            host: "6082"
            container: "6082"
          - name: vnc
            host: "5901"
            container: "5901"
          - name: audio
            host: "6100"
            container: "6100"
      env:
          - VNC_PASSWORD=dxflow
          - WALLPAPER=show
          - PANEL=show
          - TASKBAR=hide
          - AUDIO=off
          - AUDIO_PORT=6100
          - AUDIO_CHANNELS=1
          - AUDIO_RATE=22050
      resources:
          cpu: "4"
          memory: 64G
          gpu: nvidia
      link: web

[volume]
app.volume = ./volume

[port]
app.web = 6082
app.vnc = 5901
app.audio = 6100

[env]
app.VNC_PASSWORD = dxflow
app.WALLPAPER = show
app.PANEL = show
app.TASKBAR = hide
app.AUDIO = off
app.AUDIO_PORT = 6100
app.AUDIO_CHANNELS = 1
app.AUDIO_RATE = 22050

[resource]
app.cpu = 4
app.memory = 64G
app.gpu = nvidia

{
    "arch": ["amd64"],
    "image": "ghcr.io/dxflow-ai/scipion:latest",
    "version": "3.4",
    "minimum": {
        "cpu": 2,
        "memory": "32G",
        "storage": "100G"
    }
}

Notes

• Set a strong VNC_PASSWORD; it defaults to dxflow, which every reader of this page knows.
• Scipion is GPU- and compute-heavy: attach an NVIDIA GPU and give the step ample CPU, memory, and fast storage. The typical pipeline runs import → motion correction → CTF estimation → particle picking → 2D/3D classification → refinement → post-processing.
• The image installs the Scipion core; some plugins (e.g. Xmipp, RELION, MotionCor2) are installed on demand from the plugin manager and may need additional download and build time.
• The panel is kept and the taskbar hidden by default so Scipion's multiple windows are easy to manage. Set TASKBAR=show or PANEL=hide to change that.
• Audio: off by default. Set AUDIO=on to stream desktop sound; tune with AUDIO_CHANNELS (1 or 2) and AUDIO_RATE (8000/16000/22050/32000/44100). The audio port is AUDIO_PORT (default 6100) — the client follows it, so to run two sessions on one host give each its own port by setting AUDIO_PORT and the matching audio port mapping together.