Highly accurate protein structure prediction for the human proteomeNature
Determining complete protein structures is difficult as sufficient quantities need to be purified; protein size, transmembrane domains, and susceptibility to conformational changes create inconsistencies; and the work can take months to years. Tunyasuvunakool et al. applied AlphaFold to predict full-length structures with detailed chemical components of nearly the entire human proteome (98.5%) as well as ~20 other organisms and provided this data in a public database. Since publication, the authors updated AlphaFold (version 2) shortening days-long prediction computation to minutes/hours and the database now has over 200 million protein structures. [Read More]