Submitted by Anonymous on Wed, 02/12/2020 - 09:39
There has been wide interest in AlphaFold2, a program from Google’s DeepMind Artificial Intelligence (AI) network which was reported as having solved a long-standing challenge in biology. The incredible range of different behaviour and functions found across proteins depends upon their 3D structures - in turn, largely determined by protein amino acid sequences. Working out precisely how a protein chain folds from just knowing its amino acid sequence has been a major goal for decades - but now may be within reach using this AI program.
CIMR’s Prof. Randy Read commented: “The DeepMind AlphaFold team has made a huge leap forward this year. For the first time ever, computer models are nearly as good for proteins that have no obvious relatives as they are for proteins with close relatives. Advances like this don’t remove the need for experimental structural biology, but they help to focus even more of our energy on the biology itself.”
Dr Tristan Croll added: “This will be of particular benefit to experimental structural biologists working to determine the structures of previously uncharacterised proteins. Particularly in the case of x-ray crystallography it is often extremely useful to have a reliable initial model to start from using a method known as ‘molecular replacement’. In the past this has only been possible for proteins whose close relatives had already been experimentally determined - the ability of AlphaFold2 to predict the structure of brand new proteins will dramatically expand the range of situations in which this approach may be used.”
Further published details of AlphaFold2’s data are expected over the coming weeks.