Back to AI Timeline
2020Milestone

AlphaFold Solves Protein Folding

DeepMind's AlphaFold2 achieved near-experimental accuracy in predicting 3D protein structures from amino acid sequences, effectively solving a 50-year grand challenge in biology. The system outperformed all competitors at the CASP14 competition by a massive margin. This breakthrough has since accelerated drug discovery and our understanding of biological processes.

In November 2020, DeepMind's AlphaFold2 achieved a median Global Distance Test (GDT) score of 92.4 out of 100 at the 14th Critical Assessment of Protein Structure Prediction (CASP14) competition -- a level of accuracy comparable to experimental methods like X-ray crystallography and cryo-electron microscopy. The result effectively solved the protein folding problem, a grand challenge that had stumped biologists for over fifty years.

The Protein Folding Problem

Proteins are the molecular machines of life. They are composed of chains of amino acids that fold into specific three-dimensional shapes, and their shape determines their function. Understanding a protein's 3D structure is crucial for understanding how it works, how it interacts with other molecules, and how drugs can be designed to target it. However, predicting how a protein folds from its amino acid sequence alone -- the protein folding problem -- was considered one of the hardest unsolved problems in biology.

Why It Was So Hard

The number of possible configurations for even a small protein is astronomically large. Levinthal's paradox notes that a protein with 100 amino acids could theoretically adopt 10^143 different configurations. Exploring all of them would take longer than the age of the universe, yet real proteins fold into their correct shapes in milliseconds. Previous computational methods could handle small proteins but struggled with larger, more complex structures.

How AlphaFold2 Worked

AlphaFold2 used a novel neural network architecture that processed two types of information simultaneously: evolutionary relationships between related protein sequences (captured in multiple sequence alignments) and spatial relationships between amino acid pairs. The system used an iterative process where its predictions were refined over multiple passes, with each pass improving the accuracy of both the predicted distances and the predicted angles between amino acids.

The CASP14 Results

The CASP competition has been held every two years since 1994, serving as the gold standard for evaluating protein structure prediction methods. AlphaFold2's performance at CASP14 was so far ahead of all other methods that the competition's co-founder, John Moult, declared the protein folding problem "largely solved." Two-thirds of AlphaFold2's predictions were accurate enough to be useful for drug design and biological research, a threshold no previous method had approached.

The Open Science Decision

In July 2021, DeepMind published the full details of AlphaFold2 in the journal Nature and released the source code. They also partnered with the European Bioinformatics Institute to create the AlphaFold Protein Structure Database, which eventually provided predicted structures for over 200 million proteins -- essentially every known protein from every organism whose genome had been sequenced. This decision to make the technology freely available accelerated scientific research worldwide.

Impact on Biology and Medicine

AlphaFold's impact on biology has been transformative. Researchers have used it to understand disease mechanisms, identify drug targets, design new enzymes, and study the evolution of life. What previously took months or years of experimental work -- determining a single protein structure -- could now be done in minutes. The database has been accessed by over a million researchers in 190 countries.

Recognition

The achievement earned DeepMind CEO Demis Hassabis and AlphaFold lead John Jumper the 2024 Nobel Prize in Chemistry, alongside David Baker for his related work on computational protein design. It was a landmark moment for AI -- the first time an AI breakthrough received a Nobel Prize.

Key Figures

John JumperDemis HassabisJohn Moult

Lasting Impact

AlphaFold effectively solved the 50-year protein folding problem, transforming structural biology and accelerating drug discovery worldwide. It demonstrated that AI could make fundamental scientific breakthroughs, earning its creators a Nobel Prize.

Related Events

2016Milestone
AlphaGo Beats Lee Sedol

Google DeepMind's AlphaGo defeated world Go champion Lee Sedol 4-1 in a five-game match, stunning the AI community and the world. Go was considered far too complex for brute-force approaches due to its astronomical number of possible positions. AlphaGo combined deep neural networks with Monte Carlo tree search, proving AI could master intuition-heavy strategy games.

2006Research
Geoffrey Hinton's Deep Learning Breakthrough

Geoffrey Hinton and collaborators published influential work on training deep belief networks, reigniting interest in neural networks after years of stagnation. Their techniques for layer-wise pre-training made it feasible to train networks with many layers. This breakthrough is widely credited with launching the modern deep learning revolution.