Skip to content
ML Atlas

AlphaFold

DeepMind's system that predicts the 3D shape of proteins from their amino-acid sequence with near-experimental accuracy, recognized with the 2024 Nobel Prize in Chemistry.

Also AlphaFold 2 · protein structure prediction · CASP

Fig. / How it works
Amino-acid sequenceAmino-acid sequenceRelated sequencesRelated sequencesAttention trunkAttention trunkStructure moduleStructure module3D structure3D structure
Role · kind
Real systemsFamous system
Level
Intermediate

Jumper et al., DeepMind, 2020 (Nature, 2021)

A protein is a long chain that folds into a precise shape, and the shape decides what it does. Working out that shape used to take months of lab work. AlphaFold predicts it in minutes.

Where it fits

Deep Learning ArchitecturesDeep Learning ArchitecturesAttentionAttentionTransformerTransformerAlphaFold
+All relationships as a list

Is a kind of / part of

Uses / built on

See it on the full map

How it works

Proteins are chains of amino acids that fold into intricate 3D shapes, and the shape determines what a protein does. Determining shapes experimentally can take months or years, and predicting them from the sequence had been an open problem for about 50 years.

At the CASP14 assessment in 2020, DeepMind’s AlphaFold 2 predicted structures with accuracy competitive with experiments for most targets, far ahead of every other entry. Its network, built around attention layers that reason about how pairs of amino acids relate, was trained end to end on known structures (see Transformer).

DeepMind and EMBL-EBI then released predicted structures for hundreds of millions of proteins, freely available to researchers. In 2024, Demis Hassabis and John Jumper shared the Nobel Prize in Chemistry for this work (together with David Baker, honored for computational protein design).

AlphaFold is a reminder that deep learning’s impact reaches far beyond text and images.

Everyday use cases

  • Case 01

    How you've probably met it

    Through news about medicine and biology. Researchers use its predicted structures, published freely for hundreds of millions of proteins, in drug discovery and basic science.

Strengths, weaknesses and when to use it

Strengths

  • Turned a decades-old scientific grand challenge into a routine computation for many proteins.

Weaknesses

  • Less reliable for flexible or disordered regions and for some complexes; predictions still need experimental confirmation for critical uses.

When to use itAs a starting point for understanding a protein's structure and function.

Go deeper

Sources

  1. [1]Jumper et al., Highly accurate protein structure prediction with AlphaFold, Nature (2021)
  2. [2]Nobel Prize in Chemistry 2024, press release

Facts checked as of 2026-10-12