Report from CASP13

A number of our team members have been attending the CASP conferences for several years. CASP (Critical Assessment of technologies for Structure Prediction) invites researchers to submit protein sequences, and challenges the community to produce models of expected protein structure based on the sequence. For this last round (CASP13) tens of thousands of models were submitted by research groups around the world, and the best predictor of structure was made by DeepMind, an AI developed by Google. We’re hoping our structure-producing beamlines (based on actual experimental data) don’t go out of business any time soon, but are at the same time impressed by the accuracy of structure prediction that is now possible. For an in-depth analysis of the results of CASP13, see here.