AI Solved the 50-Year-Old Protein Folding Problem, Mapping 200 Million Structures

For five decades, predicting how a protein folds from its amino acid sequence was considered one of the deepest unsolved problems in biology. The answer to this question dictates how every drug works, how every disease attacks, how every enzyme catalyzes — yet the search space is so vast that, at random, finding the correct fold would take longer than the age of the universe.

In 2020, DeepMind's AlphaFold 2 ended the problem. At the CASP14 competition (the field's official benchmark), AlphaFold predicted protein structures with accuracy rivaling experimental methods like X-ray crystallography — but in seconds, not years.

The Scale of the Achievement

• 200 million protein structures released in DeepMind's free public database — essentially every protein known to science
• Equivalent to compressing decades of laboratory work into one massive open dataset
• Already accelerating drug discovery, vaccine design, and disease research
• Awarded the 2024 Nobel Prize in Chemistry

What's Next

The newest version, AlphaFold 3, predicts not just protein structures but how proteins interact with DNA, RNA, and small molecules — covering nearly the entire molecular machinery of life. Researchers are already using it to design entirely new proteins that have never existed in nature.