Drybench Cell.

What the app shows, how it works and how it was built, in the words of its own write-up, followed by the parameter defaults and the data sources with their licences.

What it shows

A guided tour of how a human cell turns a gene into a working protein, in eight steps: DNA packed in chromatin, transcription by RNA polymerase II, splicing, export from the nucleus, translation on the ribosome, the new chain in the exit tunnel, folding inside a chaperonin, and a finished protein, haemoglobin, at work.

Each step is a real structure from the Protein Data Bank, determined by X-ray crystallography or cryo-electron microscopy, with its primary paper. The colours mean the same thing in every step: DNA, messenger RNA, other RNAs, the protein machinery, the protein being made and small molecules.

How it works

The page ships the tour text and a still image for each step. The 3D view loads only when you ask for it: first the Mol* viewer (about 1.5 MB compressed), then, for each step, a gzipped BinaryCIF structure file of 25 to 240 KB (60 to 420 KB unzipped) and a MolViewSpec scene that sets the representation, colours, labels, cutaway plane and camera. The page downloads and unzips each file itself, so a download for a step you have already left is cancelled, and a stalled one ends with a message and a retry.

The structure files are cut down from the PDB entries by the build pipeline to the atoms shown, with no water or hydrogens. For the five biggest complexes by atom count, the spliceosome, TREX, the two ribosomes and TRiC, with about 36,000 to 215,000 atoms each, the machinery keeps one atom per residue and is drawn as a backbone trace, while the RNAs and chains the step is about keep every atom. That keeps them usable on a phone.

How it was built

A Python pipeline checks every PDB entry against the RCSB Data API (title, organism, method, resolution, year), checks each primary citation against Europe PMC, and checks UniProt accessions. It then reads the structures with Biotite, confirms each coloured entity is what the scene says it is, measures the numbers the text quotes with NumPy and SciPy (sizes, resolved residues, base pairs, how far the chain in the tunnel reaches from its tRNA), works out the camera and cutaway plane, and writes the trimmed, gzipped files and the scene data, noting which revision of each entry they were cut from.

The stills were rendered headless in Blender 5.2 with the Molecular Nodes add-on, using Cycles on the CPU with 32 samples and denoising. The render script reads the same scene data as the 3D viewer, so each still shows the same structure, colours, cutaway and viewing direction. The stills are the step images before the 3D view loads, and the fallback where WebGL is not available.

Parameter defaults

Structures

Eight PDB entries, trimmed to the atoms shown, with no water or hydrogens; gzipped BinaryCIF files of 25 to 240 KB.

Large complexes

The spliceosome, TREX, the two ribosomes and TRiC keep one atom per residue for the machinery, and every atom for the RNAs and chains the step is about.

Stills

Blender 5.2 with the Molecular Nodes add-on, Cycles on the CPU, 32 samples with denoising, from the same scene data as the 3D view.

Sizes are in nanometres; 1 nm is a millionth of a millimetre. The eight structures come from separate experiments published between 2006 and 2025, so they show the stages of making a protein, not one molecule passing through them. All the proteins are human. The RNA being spliced in step 3 is listed as coming from an adenovirus, and in step 5 the message is synthetic and the tRNAs have a bacterial sequence.

Data sources, licences and versions

Times on screen are local time with the UTC offset; times in exports are UTC. Dates with no time are shown at the precision the source gave them.

Today the tour has no file export. What you can take away is the address, whose ?step= link opens a step directly, and each step's text, still and primary citation, which are on the page itself.

Release gate: before the app leaves the preview, its exports and printed reports will carry the data versions, parameters, limits and a timestamp needed to reproduce a result, as the regulatory position says. This section will describe each format as it ships.

Each module below is a set of automated tests in the app's source code, with the expected results written into the tests. They check the app's logic against recorded inputs and stored reference records; none of them calls a live service.

pipeline/inside-a-cell/verified.json

The stored reference records: each PDB entry's title, method, resolution, dates and organisms from the RCSB Data API, each primary citation checked against Europe PMC, and each UniProt accession, written by pipeline/inside-a-cell/verify_sources.py.

src/components/tools/inside-a-cell/content.test.ts

Every PDB entry matches the RCSB record, every citation matches Europe PMC, the numbers in the copy are the measured ones, sizes and the colour key match the scene data, links use verified human UniProt accessions, and every step has a still within the size budget.

src/components/tools/inside-a-cell/mvs.test.ts

The palette matches the pipeline, every scene builds a MolViewSpec tree, labels can be switched off, scenes carry no camera transition, and zoom stays within limits.

src/components/tools/inside-a-cell/download.test.ts

Every shipped structure file is gzipped BinaryCIF and unzips; a stalled download fails with a timeout, a slow one is not cut off, and cancelling and server errors are reported.

src/components/tools/inside-a-cell/nav.test.ts

Step links read and write the query string, unknown steps are flagged, the pre-rendered step pages swap correctly, and the loader runs one step at a time.

Release gate: before the app leaves the preview, each release will publish a verification dataset with expected outputs on this page, so you can run your own checks.

No verification set has been published for this app yet.

Scientific computation runs in your browser, and individual apps need no account. Optional Bench accounts and backups use the site server and configured providers. The privacy statement, the hosts it contacts and the headers its workspace is served with are generated from its registry entry, the same record that sets the browser's policy.

What leaves your browser

Hosts the app contacts

The app contacts no host but its own.

Headers the workspace is served with