← The registry
№00SolvedAI × science

Predict a protein’s shape from its sequence

Verified as solved

AlphaFold 2 met the published criteria.

Read the verdict →

The problem

What needs solving

Predict the shape of a protein from the order of its building blocks, its amino-acid sequence, as accurately as a laboratory can measure it. This covers one protein chain on its own.

Why it matters

A protein’s shape decides what it does, and measuring one shape in a laboratory can take years. In 1972 Christian Anfinsen shared the Nobel Prize in Chemistry for work on how a protein’s sequence sets its shape. Since 1994 CASP, a blind contest held every two years, has asked methods to predict structures that scientists had solved but not yet published. By 2025, DeepMind says, more than 3 million researchers had used AlphaFold’s predictions of over 200 million protein structures.

What counts as progress

(a) In CASP’s blind test, a method’s median score on the global distance test, a 0-to-100 measure of how closely a prediction matches the measured shape, is about 90 or higher. (b) CASP’s independent assessors score the predictions against structures that were still unpublished when the predictions were made. (c) Solved when the method and its code are published so that others can reproduce (a) and (b).

Stater & version history

Stated by UNSOLVED editors

v1.0

18 Oct 2026 — Entered as solved for the founding registry, as the reference for what a solve looks like. It was solved before the registry opened, so the verdict is CASP’s, not the review board’s.

Explore the work on this problem