Folding

Protein structure prediction is the computational process of determining the three-dimensional shape of a protein from its amino acid sequence. The structure of a protein determines its function, so accurate predictions are essential for understanding biology and designing drugs.

Modern AI folding tools like AlphaFold 2 use deep learning models trained on millions of known protein structures. They predict inter-residue distances and angles from the sequence, then fold the chain to satisfy those constraints — achieving near-experimental accuracy for many proteins.

AlphaFold 2 uses multiple sequence alignments (MSAs) to improve accuracy by finding evolutionary information across related proteins. ESMFold skips the MSA step and folds directly from the sequence using a protein language model, making it significantly faster at some cost in accuracy.

Folding tools output PDB or CIF files containing the atomic coordinates of the predicted structure, along with per-residue confidence scores (pLDDT). These can be downloaded and visualized in the built-in PDB viewer or imported into external tools like PyMOL.

Structural biologists, drug discovery researchers, and protein engineers use folding tools to understand protein function, identify binding pockets, and guide experimental work. They are commonly used before docking or design workflows.