duckdb-zarr

field guide

duckdb-zarr
for biologists.

Explore single-cell RNA-seq and multichannel 3D fluorescence using SQL in your browser.

Starting DuckDB-Wasm…

Original datasets are fetched from their public remote sources into browser memory; SQL runs here in a Web Worker. No dataset copy is stored with the site, and no Python query server, account, or upload is needed.

01 / AnnData

Explore single cells

PBMC 68k reduced · 700 cells · 765 genes · single-cell RNA-seq

This is single-cell RNA sequencing (scRNA-seq) data: each row is one blood cell, each column a gene, and each value is that gene's measured expression in that cell. Join a cell's annotation in obs to its UMAP coordinates in obsm/X_umap, then decide which cells to show.

Sample data: Scanpy PBMC 68k reduced ↗, a processed subset of 10x Genomics blood-cell data from Zheng et al. (2017) ↗.

AnnData guide · Loading provenance…

Plot columns: x, y, label. Other result shapes are shown as a table.

Waiting for data…

What am I querying?

pbmc_obs reads the annotations; pbmc_coordinates reads the two embedding components. These are SQL views over Zarr, not precomputed JSON points. The source already contains this UMAP; the browser uses it unchanged. A UMAP axis is an embedding coordinate, not a physical distance. cell_type (aliasing the source's bulk_labels) comes from matching each cell's expression profile to reference bulk RNA-seq expression from sorted immune-cell populations — it is an annotation derived from the data, not ground truth from a separate assay. Point "Data source" at another AnnData Zarr store with the same obs/obsm/X_umap layout (e.g. hosted on S3/HTTPS with CORS enabled) to explore it instead; the query views are rebuilt against whatever URL you load.

02 / OME-Zarr

Explore a 3D fluorescence volume

Loading dataset…

Both fluorescence channels are loaded together: blue marks LaminB1 and yellow marks Dapi. Choose a depth range or isolate one channel to explore it. The volume turns gently on its own; drag to explore, scroll to zoom, or pause the rotation for a closer look.

Sample data: IDR image 6001240 (idr0062) ↗ from the NesSys study by Blin et al. (2019) ↗ · Dataset DOI ↗ · CC BY 4.0 ↗.

OME-Zarr guide · Loading provenance…

Plot columns: c, x, y, z, value. Channel colors and physical spacing come from the source metadata. Automatic display contrast suppresses background haze without changing SQL intensities. Non-image query results are shown as a table.

Waiting for data…

What am I querying?

Loading provenance…