AlphaFold2 — Deep Learning Breakthrough in Protein Structure PredictionNew

AlphaFold2 (Jumper et al., Nature 2021) is DeepMind's end-to-end deep learning model that predicts a protein's 3D structure directly from its amino-acid sequence. At the CASP14 assessment it reached near-experimental (atomic) accuracy for the first time, effectively solving a 50-year-old grand challenge in biology.

The open release of both inference code (Apache-2.0) and model parameters (CC-BY-4.0), together with the AlphaFold Database of millions of predicted structures, has reshaped structural biology, drug discovery and enzyme design. A follow-up, AlphaFold3 (2024), extended the framework to joints over proteins, nucleic acids, ligands and more.