Skip to content
YAC
The YAC app with a conversation on the left and a dashboard of linked charts about a clinical cancer cohort on the right.

YAC

Ask questions of biomedical datasets in plain language. YAC answers with linked, interactive charts that you can refine, filter and save.
Try the demo sessionRead the docs

Pick a dataset

Section titled “Pick a dataset”

Open a predefined session to see what YAC can do without an API key, or start a new session and ask your own questions.

Clinical cancer cohort (pcx)

De-identified clinical records for 970 patients across 7 linked tables: demographics, disease events, medical therapy, radiation, surgery, survival and treatment summary.

  • Cohort tour: Age at diagnosis, treatment sequencing, survival by diagnosis cohort, resection extent, radiation sites and disease events.
  • Dashboard overview: Patients by data cohort and diagnosis cohort, and by organization.
New session

Survival cubes (pcx)

Pre-aggregated event-free and overall survival counts for 970 patients with ATRT or medulloblastoma, by diagnosis, treatment protocol, molecular subtype and site.

New session

Seattle records + survival cubes (pcx)

Line-level clinical records for the 93 radiant-cohort patients at Seattle Children's Hospital, alongside the event-free and overall survival cubes for all 970 patients, linked by diagnosis, protocol, subtype and site.

New session

HuBMAP

Donor, sample and dataset metadata from the Human BioMolecular Atlas Program portal.

New session

Palmer penguins

Body measurements for 344 penguins of three species: a small, single-table dataset to try YAC on.

New session

How it works

Section titled “How it works”

Ask

Type a question about the data. The agent chooses a chart from a library of templates, binds it to your dataset’s tables and fields, and adds it to the dashboard.

Explore

Charts are linked. Brush a range or click a bar to filter every other chart. Swap the field a chart splits by from the dropdowns under it in the chat.

Keep

Save a session in your browser, or export it as a file to share. A session brings back the conversation and the dashboard.

Bring it home

The chat is an npm package (udi-yac) and the agent a Python one (udiagent). See Integrate.

Cite YAC

Section titled “Cite YAC”

Lange D, Gao S, Sui P, Misner P, van den Brandt A, Money A, Akhmetov N, Choy L, Zitnik M, Gehlenborg N. YAC: Bridging Natural Language and Interactive Visual Exploration with Generative AI for Biomedical Data Discovery. arXiv:2509.19182, 2025. doi:10.48550/arXiv.2509.19182

@misc{lange2025yac,
  title         = {YAC: Bridging Natural Language and Interactive Visual Exploration with Generative AI for Biomedical Data Discovery},
  author        = {Lange, Devin and Gao, Shanghua and Sui, Pengwei and Misner, Priya and van den Brandt, Astrid and Money, Austen and Akhmetov, Nikolay and Choy, Lisa and Zitnik, Marinka and Gehlenborg, Nils},
  year          = {2025},
  eprint        = {2509.19182},
  archivePrefix = {arXiv},
  primaryClass  = {cs.HC},
  doi           = {10.48550/arXiv.2509.19182},
  url           = {https://arxiv.org/abs/2509.19182}
}