AlphaFold at Five: 3 Million Researchers, a Nobel, and a New Playbook for AI-Driven Science

Five years after AlphaFold 2 solved the protein structure prediction problem at CASP 14 in 2020, the system has been used by more than 3 million researchers across 190-plus countries and earned the 2024 Nobel Prize in Chemistry — the clearest proof yet that AI can function as a core instrument of science rather than a laboratory curiosity.

In a retrospective published by the AlphaFold team — including Demis Hassabis, John Jumper, Pushmeet Kohli and Anna Koivuniemi — DeepMind traces how a competition-winning model became what it describes as "a tipping point toward AlphaFold becoming a scientific tool adopted around the world."

The stakes are hard to overstate. Proteins are the microscopic machines that drive every process in a living cell, folding from chains of amino acids into 3D shapes that define their function. Knowing those shapes is critical for drug discovery and understanding disease. Before AlphaFold, determining a single structure could take a year or more of expensive experimental work. Predicting structure from amino acid sequence alone had stood as a 50-year-old grand challenge in biology.

AlphaFold 2 cracked it at the CASP 14 (Critical Assessment of protein Structure Prediction) competition, predicting structures with what the team calls "astonishing accuracy." But the lasting impact, the team argues, came from open access rather than the model itself.

From model to infrastructure

In 2021, DeepMind launched the AlphaFold Protein Database in partnership with EMBL-EBI. One year later it released predictions for more than 200 million protein structures — output the team says would have taken "hundreds of millions of years to solve experimentally." The database is freely available, and more than 1 million of its users are in low- and middle-income countries. Over 30% of AlphaFold-related research focuses on understanding disease.

The adoption numbers are the strongest evidence of scale. AlphaFold has been cited in more than 35,000 papers, and more than 200,000 papers incorporated elements of AlphaFold 2 in their methodology, according to the team.

An independent analysis by the Innovation Growth Lab quantifies the shift: researchers using AlphaFold 2 show an increase of over 40% in their submission of novel experimental protein structures. Those structures are more likely to be dissimilar to known structures. Research linked to AlphaFold 2 is twice as likely to be cited in clinical articles, and significantly more likely to be cited by a patent, than typical structural biology work.

From honeybees to heart disease

The retrospective highlights concrete applications. European scientists used AlphaFold to understand Vitellogenin (Vg), a key immunity protein in honeybees; those structural insights now guide conservation efforts for endangered bee populations and AI-assisted breeding programs.

In cardiology, AlphaFold 2 helped reveal the structure of apolipoprotein B100 (apoB100), the central protein in LDL, or "bad cholesterol," whose shape had eluded researchers for decades. Atherosclerosis is the leading cause of global mortality, and the newly mapped cage-like structure gives pharmaceutical researchers atomic-level detail needed to design preventative heart therapies.

The team also points to cases of democratized access. Turkish undergraduate students Alper and Taner Karagöl taught themselves structural biology during the pandemic using online AlphaFold tutorials, with no prior training. They have since published 15 research papers. Cyril Zipfel, professor of Molecular & Cellular Plant Physiology at the University of Zurich and the Sainsbury Lab, used AlphaFold alongside comparative genomics to understand how plants perceive environmental changes, work the team says paves the way for more resilient crops.

One of the most-viewed structures in the database is p53, a cellular tumor antigen related to cancer.

Isomorphic Labs and the road to AlphaFold 3

The commercial trajectory matters as much as the academic one. Isomorphic Labs, the AI drug discovery company founded in 2021, was created when the breakthrough model proved powerful enough for rational drug design. The company has since built a unified drug design engine, with what the team describes as an ambition "to one day solve all diseases."

Together with Isomorphic Labs, DeepMind developed AlphaFold 3, which extends prediction beyond proteins to DNA, RNA and ligands — the small molecules that make up most drugs. It can generate joint 3D structures of entire molecular complexes, showing how a candidate drug molecule binds to its target protein or how proteins interact with genetic material. The AlphaFold Server, available to non-commercial researchers, has produced more than 8 million fold predictions for thousands of researchers worldwide.

John Jumper, described by the team as one of the most instrumental scientists behind AlphaFold, has discussed the shift in architecture from AlphaFold 2 to the broader AlphaFold 3 and the tool's unexpected applications since his Nobel win.

A template, not a peak

DeepMind frames the next generation of models as direct descendants of the AlphaFold playbook. AlphaMissense and AlphaGenome assess the genetic mutations underlying disease. AlphaProteo designs novel, high-strength protein binders targeting molecules associated with cancer and diabetes.

"Biology was our first frontier, but we view AlphaFold as the template for how AI can accelerate all of science to digital speed," the team writes, naming fusion, Earth sciences and scientific discovery broadly as targets for the next AlphaFold-like breakthroughs.

That framing signals where this story goes next: DeepMind is positioning AlphaFold not as a completed achievement but as the reference case for AI-accelerated science — a claim the next five years of drug pipelines and disease research will test.