AlphaFold Is Five Years Old – These Charts Show How It Revolutionized Science (www.nature.com)

Five years after DeepMind unveiled AlphaFold2 in November 2020, the AI system has reshaped structural biology by delivering near-experimental accuracy for protein 3D models and making those models widely accessible. DeepMind released the code and parameters, enabling labs to run predictions at scale and fueling the AlphaFold database (hosted by EMBL–EBI), which now holds more than 240 million predicted structures and has been accessed by roughly 3.3 million users across 190+ countries (including over a million users in low- and middle-income countries). The 2021 AlphaFold paper has been cited in about 40,000 articles, and its developers have received major recognition for the work. Technically, AlphaFold2 was trained on Protein Data Bank (PDB) entries and produces models that often resolve ambiguities in X‑ray crystallography and cryo‑EM maps, speeding hypothesis generation and experimental design. Practical impacts include accelerating discoveries such as the identification of a sperm–egg interaction complex (Tmem81 and Bouncer) and an observed ~50% increase in PDB submissions by researchers who used AlphaFold versus a non‑using baseline in a recent impact study. The model’s combination of high accuracy, open access, and scalability has democratized structure prediction, boosted throughput for experimental structural biology, and integrated computational prediction more tightly into biological and drug‑discovery workflows.