STEAM-v1 cross-species enhancer view (Fig 6e explorer)

Interactive recreation of Fig 6e from Shendure et al., evolutionary transfer learning, for arbitrary hg38 coordinates. Enter a position and a cell type; the app pulls the hg38-projected STEAM-v1 predicted-accessibility track from all 239 Zoonomia mammals, orders species by the Zoonomia 241-species phylogeny, and renders a viridis heatmap with synteny-coverage and summed-strength tracks above.

Layout

• Top-left: STEAM-v1 score color bar.
• Top tracks: per–100 bp bin, fraction of species with hg38 coverage (gray) and per-column normalized summed STEAM-v1 score across species (teal).
• Left: rectangular phylogram from the Zoonomia HAL guide tree, ladderized, with italic abbreviated tip labels (red for the Highlight species list).
• Main: viridis heatmap of predicted accessibility per species × 100 bp bin, NA cells set to the viridis floor (#440154).

Run locally

pip install -r requirements.txt
streamlit run app.py

The first query for a fresh locus fetches ~239 remote bigwigs (~30–60 s with 32 parallel workers); repeats are instant from the in-memory cache.

Deploy to Streamlit Community Cloud

1. Push this repo to GitHub.
2. Sign in at https://share.streamlit.io with your GitHub account.
3. Click New app → pick this repo + branch + app.py → Deploy.

The free tier (Community Cloud) has 1 GB RAM, which fits ~20 cached loci. No secrets or extra config are required; everything the app needs is fetched on demand.

Data sources

• Per-species hg38-projected STEAM-v1 tracks: https://shendure-web.gs.washington.edu/content/members/cxqiu/public/nobackup/jax_atac_augmented_241_mammals_hg38/
• Per-species native-coord enhancer BEDs (not used here; available at .../jax_atac_augmented_241_mammals_core_enhancer/).
• Zoonomia 241-mammal phylogeny (data/zoonomia_241.nwk): extracted via halStats --tree from the 241-mammalian-2020v2.hal Cactus alignment. All 239 species in the per-species set match a tree leaf exactly.

Caveats

• The synteny track here is a coverage proxy from the hg38 projection, not a formal chain/HAL-derived span. Swap in real spans by editing fig6e_core.compute_synteny if you have a chain-based table.
• Mus_musculus and Homo_sapiens are the model references and are not in the 239-species per-species set; if you highlight them they're silently ignored.
• Fig 6d (synteny network) and Fig 6f (per-cluster stratification) need the master enhancer-call synteny graph and are not implemented here.

Project layout

.
├── app.py                  # Streamlit UI
├── fig6e_core.py           # core fetch / synteny / tree / plot (no Streamlit dep)
├── build_notebook.py       # builds fig6e_interactive.ipynb (notebook variant)
├── fig6e_interactive.ipynb # standalone notebook variant of the same logic
├── data/
│   └── zoonomia_241.nwk    # Zoonomia 241-mammal tree (Newick, ~12 KB)
├── requirements.txt
└── .streamlit/config.toml