AlphaFold

AlphaFold is an artificial intelligence system that predicts a protein’s three-dimensional structure from its amino-acid sequence. Its predictions have made structural models available for proteins that have not been experimentally solved.

AlphaFold 2: The 2021 version of AlphaFold that achieved highly accurate protein-structure predictions in the CASP14 assessment. Its architecture and results established the system’s best-known form.
AlphaFold Protein Structure Database: A public database of protein-structure predictions generated by AlphaFold for many organisms. It distributes large-scale predictions beyond the proteins studied in the original benchmark.
X-ray crystallography: A method for determining molecular structures from X-ray diffraction by crystals. It measures structural evidence experimentally rather than predicting a structure from sequence.
AlphaFold 3: A model introduced in 2024 for predicting structures and interactions among proteins, nucleic acids, small molecules, ions, and modified residues. It extends AlphaFold’s scope from protein structures toward interacting biomolecular systems.
Multiple sequence alignment: An arrangement of related biological sequences that highlights conserved and variable positions. Correlated changes across aligned protein sequences provide evidence about residue contacts.
Protein structure prediction: The computational task of inferring a protein’s three-dimensional structure from sequence or related information. AlphaFold is one influential system within this broader task.
Nuclear magnetic resonance spectroscopy: A method that uses nuclear magnetic resonance signals to determine molecular structures and dynamics, often in solution. It can reveal experimentally observed protein conformations and dynamics absent from a single prediction.
Protein Data Bank: The worldwide archive of experimentally determined three-dimensional structures of biological macromolecules. Its curated experimental structures supplied training data and remain a reference for evaluating predictions.
Attention (machine learning): A neural-network operation that weights relationships among elements in a sequence or other input. Attention layers help AlphaFold relate distant amino-acid positions.
Protein design: The creation or modification of amino-acid sequences to produce proteins with specified structures or functions. Predicted structures can help researchers assess designed sequences.

Linked from 7 pages