Collaborative data analysis using Jobs

This guide explains how to work with colleagues on the same dataset using Projects and Jobs: how to share access, find each other’s processing runs, rerun workflows, and keep your work organized.

What gets shared when you collaborate

When you share a project with someone, they get access to:

• Raw data and uploaded files in the project
• Processed outputs generated by jobs (in the project folders)
• The job history for that project (jobs run before and after they join)

That means your teammates can review your analysis without exchanging files manually. Dependent on their roles they can also reproduce and improve on it directly.

Roles

Projects can be shared with four different roles. Each role includes everything the roles below it can do, plus additional permissions.

• Owner: Full control. Manages all roles and is ultimately responsible for the project. There is exactly one Owner per project, typically the PI or project lead. The project creator becomes Owner by default but can transfer ownership.
• Manager: Operational management. Can add or remove Members, request deletions of raw data, approve deletion of non-raw data, and manage linked datasets.
• Member: Hands-on contribution. Can upload data, request deletion of non-raw data, run and rerun jobs, edit the project logbook and add Viewers. Cannot manage Managers or request deletions of raw data. Best for PhD students, visiting scientists, and active collaborators.
• Viewer: Read-only access. Can browse files and view job details, but cannot run analysis or modify anything. Best for passive collaborators, reviewers, or anyone who needs visibility without write access.

Use Managers sparingly to keep access controlled.

Step 1 : Share a project with your team

1. Go to Projects
2. Select the project
3. Click on Manage Access
4. Add people via their username or e-mail adress and assign them their intended role

Tip: Add at least one other Manager as backup (vacations, handover).

Step 2 : Find your team’s jobs and outputs

Option A: Start from Jobs

1. Go to Analysis → Jobs
2. Use the available filters/search:
3. • Search by triggerer (the person who ran the job)
   • Search by job template name (e.g., “Render Molecule”, “DIALS”, etc.)
3. Open a job to view:
4. • Input files
   • Parameters used
   • Logs
   • Output files
   • Comments

Option B: Start from the project file space

1. Open the Project
2. Go to the Processed (or output) folder
3. Each job typically creates its own output folder
4. Browse results (images, logs, reports, output datasets)

Use this when the job list is long and you want to quickly find “what was produced”.

Step 3 : Review a teammate’s job

For any job you can usually inspect:

• Job Details (inputs, parameters, machine type)
• Output files
• Logs (useful for troubleshooting)
• Comments (if you have a role of member or above)

This is the fastest way to understand what a colleague did and whether the result looks correct.

Step 4 : Rerun a teammate’s job to iterate

Rerunning is the recommended way to collaborate:

• You keep the same workflow and starting point
• You can adjust parameters, inputs, or performance settings
• The rerun creates a new output folder (so you won’t overwrite anyone’s results)

How to rerun

1. Open the job in Jobs
2. Click Rerun
3. Review the step-by-step flow:
4. • Inputs (verify file types are correct)
   • Parameters (tune for improved results)
   • Machine type/performance (CPU vs GPU, turbo vs standard)
4. Start the job

Result: Your rerun will appear in the job list with you as the triggerer.

Step 5 : Coordinate inside the project

Use the Project Logbook

Each project has a Logbook where the team can write:

• Why someone was invited
• What has been processed and what still needs doing
• Which job IDs correspond to key results
• Notes about “good parameter sets”

Suggested logbook template:

• Goal / Dataset
• Latest best job(s): Job ID(s) + short reason
• What changed: parameters / inputs
• Next steps / open questions
• Owner & date

Add comments to jobs

Use job comments to capture:

• “This run used GPU for the movie render”
• “Fixed wrong input file type”
• “Best output: view=all + cartoon style”

Best practices for smooth collaboration

• Name projects clearly (dataset + date + purpose)
• Use consistent job templates across the team
• Record “best run” job IDs in the logbook

Use output folders as the “source of truth” for results, and jobs as the reproducible history