AlphaFold

AlphaFold is a deep learning system developed by DeepMind that predicts the three-dimensional structure of proteins from their amino acid sequences. The original AlphaFold (2018) used neural networks trained on known protein structures; AlphaFold2 (2020) achieved a major breakthrough by reaching near-experimental accuracy on the CASP14 protein-folding competition, solving a problem that had resisted solution for decades.

The system uses a transformer-based architecture with multiple sequence alignment and structural templates to predict atomic coordinates. Its success demonstrated that machine learning could solve fundamental problems in structural biology that previously required years of experimental work via X-ray crystallography or cryo-EM.

DeepMind released AlphaFold2 as open-source software in 2020, enabling rapid adoption across academia and industry. The system has been used to predict structures for millions of proteins, including the entire human proteome and proteins from organisms of medical importance, accelerating drug discovery and biological research.