ProtSpace Web is a browser-based visualization tool for exploring protein language model (pLM) embeddings. Built with modular web components (canvas renderer, interactive legend, control bar), it enables interactive exploration through dimensionality reduction methods (PCA, UMAP, t-SNE) with zoom, pan, and selection. Color by annotations, view 3D protein structures, and export images or data files for sharing.
Demo: https://protspace.app/ → Drag & drop .parquetbundle files
Option 1: Google Colab (no local installation needed)
Generate .parquetbundle files directly in your browser:
Option 2: Python ProtSpace (local installation)
pip install protspace
# Query UniProt and generate visualization files
protspace-query -q "(ft_domain:phosphatase) AND (reviewed:true)" -o output_dir
# Or use your own embeddings
protspace-local -i embeddings.h5 -o output_dirSee the Python ProtSpace repository for details.
Full Documentation - User guides, data preparation, and feature explanations.
git clone https://github.com/tsenoner/protspace_web.git
cd protspace_web
pnpm install
pnpm dev # App: http://localhost:8080 | Docs: http://localhost:5174/docs/Before committing, run:
pnpm precommitThis matches the installed local Git hook by running lint-staged, repo-wide type checks, Knip,
dependency-hygiene checks, the local unit/integration test suite, and a docs build.
For a faster static-only pass while you are iterating, run:
pnpm qualityApache License 2.0 - see LICENSE