Protein Provenance Card

Seven questions to ask of any protein before you design with it. About 15 minutes, with three free databases.

Raphael Kim· 13 Nov 2026· v1.0

How to use it

Pick one protein that matters to your work: the one in your mycelium, your glue, your dye, your fibre. Open the three databases below in separate tabs. Then go down the card and write a short answer to each question.

A blank is an answer too. Writing “not recorded” tells you something about how much we really know.

UniProt → Protein Data Bank → AlphaFold Database →

To fill it in on paper, print this page or save it as a PDF from your browser.

The card

Which protein, from which organism?

Look

UniProt. Search the protein name and the organism together.

Write

The name, the organism (common and Latin), and the UniProt entry number.

How well checked is the entry?

Look

The top of the UniProt entry. Reviewed means a person checked it by hand. Unreviewed means it was added automatically. Then find the line called “Protein existence”: evidence at protein level, evidence at transcript level, inferred from a related species, predicted, or uncertain.

Write

Reviewed or unreviewed, and the evidence level.

Has anyone measured its shape?

Look

The Protein Data Bank. Search with the UniProt entry number. The “Structure” section of the UniProt entry lists measured shapes too.

Write

The PDB ID, the method (X-ray, cryo-EM or NMR), the year and the lab. If only a relative was measured, write which organism it came from.

What does the prediction say, and how sure is it?

Look

The AlphaFold Database. Search with the UniProt entry number. The colours show confidence, part by part: dark blue is very high (a score over 90), light blue is confident (70 to 90), yellow is low (50 to 70), orange is very low (under 50). Orange parts are often loose and don’t hold one shape at all.

Write

Which parts the model is sure about, and which it isn’t.

What can’t the shape show?

Look

The “Function” and “PTM / Processing” sections of the UniProt entry. Look for parts the organism cuts off, sugars it attaches, chemical changes it makes after the protein is built, and partners the protein needs.

Write

One or two things that matter for your project and that no shape on a screen will show you.

Where did the organism come from, and who was there?

Look

The papers listed in the UniProt entry, the record of the strain in its culture collection, and any notes on sharing the benefits of genetic resources. See digital sequence information and biocultural labels for why this matters.

Write

The place and the year, if you can find them, and anything about the people who live or work there. If you can’t find them, write “not recorded”.

Your tag, and what you’d need to believe

Tag

Measured (give the PDB ID), Predicted (give the model and how confident it was), or Unknown (no shape at all yet).

Write

One line: what would you need to believe about this protein to design with it?

A worked example: my mussel glue

My own project is a gentle glue for mending heritage objects, using a protein a mussel uses to hold on to wet rock. Here is its card.

Mussel foot protein 5 (mfp-5), from the Mediterranean mussel, Mytilus galloprovincialis. UniProt Q6QZR3.

Unreviewed. Evidence at transcript level only, and listed as a fragment.

Not measured. A close relative from the Asian green mussel has been: PDB 7QAB.

Mostly low confidence. The protein stays loose and doesn’t settle into one shape.

The stickiness. The mussel turns some of the protein’s building blocks into a sticky form called DOPA, in its own body. The prediction doesn’t show that step.

Not found yet. I’m still looking.

Predicted, low confidence
To design with it, I’d need to believe the glue works through its chemistry more than its shape. That’s something to test, not assume.

Share yours. Reply to any Biodesign Academy email with your card, or just the organism and the tag. Organisms that matter to biodesign but have no measured proteins go on a list I’m building, and I’ll share it in the newsletter.

This card is the week 2 exercise from my course, AI-Native Biodesign: With Proteins. It’s the judgement part of being an AI-native biodesigner: reading what a database can show you, and what it can’t.