Protein Structure & Drug Discovery

AlphaFold2 (2020, CASP14) collapsed a 50-year experimental bottleneck — predicting 3D protein structure from amino-acid sequence — into a compute problem, at near-experimental accuracy, scaled to ~200M proteins.

What changed: structural hypotheses became free; a lab can now get a plausible structure for anything. What didn't: a predicted structure is a static snapshot — proteins move, bind partners, get modified — and knowing the shape of a lock is step one of drug design, not the drug. The bottleneck moved downstream, not away.