hide
Free keywords:
-
Abstract:
I was the assessor for high-accuracy models in last year CASP experiment (Critical Assessment of Structure Prediction), in which computational groups try to predict the three-dimensional structure of proteins for which neither they nor the organizers know the answer at the time of prediction. For the first time in CASP history all targets ended up being high-accuracy targets, because one predictor (AlphaFold2) predicted structures at an accuracy that often rivaled that of experimental methods. After being actively engaged in CASP for over two decades and seeing essentially no progress between 2002 (CASP5) and 2016 (CASP12), I had given up hope that the protein folding problem would be breached in my lifetime, but now it has. I will cover as far as I can the main questions I am constantly asked: What happened at CASP14? Are AlphaFold2 predictions as good as crystal structures? Is the protein folding problem now solved? Does AlphaFold2 predict protein interactions? Dynamics? Unstructure? Where do the other leading groups stand (Baker, Zhang, Tencent)? I will also try to outline some thoughts about how this will impact the Life Sciences.
