proteinfamilies: feedback on its Depictio dashboard template

[weber8thomas]

Hi proteinfamilies team and users,

Depictio is a generic, open-source platform developed at EMBL, an alternative to the custom Shiny or Streamlit app each group ends up writing to explore pipeline results, and usable as a companion web app for a publication. A pipeline gets a dashboard template: a short, shareable definition that turns the outputs of any run into an interactive dashboard, in the spirit of what shinyngs does for expression data. It is built around MultiQC and adds the downstream analysis files and other data types next to it.

For proteinfamilies: the dashboard shows how the input sequences cluster into families, family size and conservation, how close the members are to their family representative, the family alignment and the profile HMM logo, next to the MultiQC report.

💡 Why a template?

• No app to maintain: the dashboard is a definition, reused on every run.
• MultiQC reports (aggregated across runs), result tables and other data types in one place, with linked filters.
• Fully customisable from the GUI or in YAML, per dashboard.
• Python package, Docker Compose or Kubernetes. Readers just open a link.
• Versioned dashboards and data, with comments and annotations on any component.

📦 Which data can it show?

Implemented:

• MultiQC reports, aggregated across runs
• Tables (CSV, TSV, Parquet, Feather, Excel)
• Images (PNG, JPEG, TIFF...)
• Geographic data (GeoJSON, lat/lon tables)
• Phylogenetic trees (Newick, Nexus)
• Protein structures, alignments and sequences (PDB, mmCIF)

Ongoing (open pull requests):

• Bioimages (OME-Zarr, including SpatialData stores, and OME-TIFF)
• Genomic files in JBrowse 2 (BAM, CRAM, VCF, BigWig, BED, GFF...), linked to the dashboard filters

👀 Try it for proteinfamilies

Open the proteinfamilies dashboard (no account needed to browse).

🙏 What we need

I have developed 28 templates so far and have already exchanged with some developers of ampliseq, viralrecon and variantbenchmarking. They now need review from people who actually run these pipelines, to assess and criticise the dashboards. For proteinfamilies, we need the view of its developers and of the people who use its results: you know what people look at first and what would mislead them.

Each template has a status level (Draft, Experimental, Reviewed, Certified), described in the docs. The proteinfamilies template is currently: Draft.

💬 How to give feedback

• Sign in with Google, then comment on the whole dashboard or on a single component.
• Or just reply in this thread.

✅ Status

• Feedback from proteinfamilies users
• Feedback from proteinfamilies developers
• Feedback triaged
• Changes released

Comments are only used to improve the dashboard.