Protein Folding
The process by which a protein's amino acid sequence determines its three-dimensional structure — a 50-year unsolved problem that AlphaFold largely solved in 2020 using deep learning.
What it means
Protein folding is the physical process by which a newly synthesized chain of amino acids spontaneously adopts its functional three-dimensional structure. The sequence of amino acids (the primary structure) determines the final folded shape (the tertiary structure) through a combination of hydrogen bonding, hydrophobic interactions, van der Waals forces, and electrostatic interactions.
Understanding a protein’s structure is critical because function follows form: an enzyme’s active site, a receptor’s binding pocket, and an antibody’s specificity all depend on the precise geometry of the folded protein.
Why it was a hard problem: The number of possible conformations for even a modest protein is astronomically large (Levinthal’s paradox). Experimental methods — X-ray crystallography, cryo-EM, NMR — can determine structures, but are slow, expensive, and not always feasible for every protein. For decades, predicting structure from sequence computationally was considered one of biology’s grand challenges.
The AlphaFold moment
In 2020, DeepMind’s AlphaFold 2 achieved near-experimental accuracy in the CASP14 competition, effectively solving the protein structure prediction problem for most protein families. The AlphaFold Protein Structure Database now holds predicted structures for over 200 million proteins across nearly all known organisms.
What “solved” means (and doesn’t mean):
- AlphaFold predicts static, single-chain structures with high accuracy for proteins with known homologs
- It does not predict protein dynamics, disordered regions, or how a protein changes conformation when it binds a ligand
- It performs less well on proteins with no evolutionary relatives (orphan proteins) and on structural complexes without additional tools
- AlphaFold3 (2024) extended predictions to protein-DNA, protein-RNA, and protein-small-molecule complexes
Related tools
- AlphaFold — structure prediction from sequence
- ESMFold — faster single-sequence prediction
- RFdiffusion / ProteinMPNN — designing proteins with desired structures