Protein structure prediction

Computational inference of a protein’s three-dimensional shape from its amino-acid sequence; CASP benchmarks blind-test methods against unpublished experimental structures.

Protein structure prediction asks what three-dimensional shape a chain of amino acids will fold into. That shape largely determines biological function. For decades the problem resisted purely computational solutions; community CASP competitions score methods on sequences whose experimental structures stay hidden until evaluation.

AlphaFold 2 at CASP14 (article) brought deep learning accuracy close to slow laboratory techniques for many targets, turning AlphaFold from a research prototype into shared infrastructure.