\n<\/p>\n
That meant making a database with all the capabilities would have been impossible. Instead, we\u2019ve released AlphaFold Server, a free tool that lets scientists plug in their own sequences that AlphaFold can then generate molecular complexes for. Since launching in May, researchers have already used it to generate over 1 million structures.<\/p>\n
\u201cIt\u2019s like Google Maps for molecular complexes,\u201d says Lindsay Willmore, research engineer at Google DeepMind. \u201cAny user who doesn’t know how to code at all can just copy and paste the sequences of their proteins, DNA, RNA or the name of their small molecule, press a button and wait a few minutes. Their structure and the confidence metrics will come out so that they’re able to look at and evaluate their prediction.\u201d<\/p>\n
In order to get AlphaFold 3 to work with this much wider range of biomolecules, the team vastly expanded the data that the newer model was trained on to include DNA, RNA, small molecules and more. \u201cWe were able to say, \u2018Let’s just train on everything that exists in this dataset that helped us so much with proteins and let\u2019s see how far we can get,\u2019\u201d Lindsay says. \u201cAnd it turns out we can get pretty far.\u201d<\/p>\n
Another major change in AlphaFold 3 is a shift in architecture for the final part of the model that generates the structure. Where AlphaFold 2 used a complex custom geometry-based module, AlphaFold 3 uses a generative model that\u2019s based on diffusion \u2014 similar to our other cutting-edge image generation models, like Imagen \u2014 which greatly simplified how the model handles all the new molecule types.<\/p>\n
That shift led to a new issue, though: Since so-called \u201cdisordered regions\u201d of proteins weren\u2019t included in the training data, the diffusion model would try to create an inaccurate \u201cordered\u201d structure with a defined spiral shape, instead of predicting disordered regions.<\/p>\n
So the team turned to AlphaFold 2, which is already extremely good at predicting which interactions would be disordered \u2014 which look like a pile of chaotic spaghetti \u2014 and which ones were not. \u201cWe were able to use those predicted structures from AlphaFold 2 as distillation training for AlphaFold 3, so that AlphaFold 3 could learn to predict disorder,\u201d Lindsay says.<\/p>\n
\u201cWe have a saying: \u2018Trust the fusilli, reject the spaghetti,\u2019\u201d adds Jonas.<\/p>\n<\/div>\n