Protein structure prediction is the computational determination of a protein’s three-dimensional folded structure from its amino-acid sequence. Deep-learning systems such as AlphaFold, built on attention-based architectures, achieved near-experimental accuracy and transformed structural biology. It is a landmark AI application with deep impact on drug discovery, molecular biology, and rational enzyme design.
Content
- Modern systems combine multiple-sequence-alignment features with attention-based networks that reason jointly over residue pairs and geometry, predicting inter-residue distances and orientations and refining full atomic coordinates. The resulting structures, released at proteome scale, accelerate hypothesis generation in biology, enable structure-based drug design, and have extended to predicting complexes and protein-ligand interactions.