AlphaFold 2

Summary

AlphaFold 2 predicts three-dimensional protein structures with near-experimental accuracy for many targets. Its CASP14 performance established deep learning with evolutionary and structural reasoning as a practical route to large-scale protein structure prediction.

Method

The model uses multiple-sequence alignments and a learned representation of residue relationships, then iteratively refines coordinates and confidence estimates. It was designed for protein monomers and does not natively cover the full range of biomolecular interactions supported by AlphaFold 3.[1]

Downstream Use

• DrugCLIP uses predicted protein structures as part of a rapid virtual-screening workflow for protein pockets and small molecules.

Citations

[1] Jumper et al. (2021), "Highly accurate protein structure prediction with AlphaFold"