For fifty years, going from a protein’s sequence to its 3D shape was biology’s great unsolved puzzle. Today’s digest is about the revolution that cracked it — and turned structure into something you can look up. Jumper et al.’s AlphaFold arrived at atomic accuracy because ‘proteins are essential to life, and understanding their structure can facilitate a mechanistic understanding of their function.’ Baek et al.’s open-source RoseTTAFold matched it with a ’three-track network,’ even building complexes ‘from sequence information alone, short-circuiting traditional approaches that require … docking.’ Tunyasuvunakool et al. folded a whole proteome — noting that ‘after decades of effort, 17% of the total residues in human protein sequences are covered by an experimentally determined structure.’ Varadi et al.’s AlphaFold DB made it open — ‘over 360,000 predicted structures’ heading toward ‘over 100 million.’ Lin et al.’s ESMFold did it from a ’large language model’ with an ‘order-of-magnitude acceleration,’ folding ‘>617 million metagenomic protein sequences.’ And Abramson et al.’s AlphaFold 3 went beyond the monomer to whole ‘biomolecular interactions.’ A structure for (almost) every sequence.
