AlphaFold 2 structure modelling: access, uses and limitations
Time: 3:00 PM – 4:00 PM BST
At this webinar we will hear from Professor Dan Rigden from the Department of Biochemistry and Systems Biology, University of Liverpool.
AlphaFold 2 (AF2) is a deep learning network that enables protein structure prediction of unprecedented accuracy. Since the code was released Open Source, models are readily accessible using suitable workstations and even using online resources. The AlphaFold database also provides a large and growing collection of pre-calculated models. Whatever their origin, and with appropriate care taken in model interpretation, AF2 models can substitute for experimental structures in, for example, the planning of lab experiments and interpretation of the results in a structural context. They also enable better annotation of protein function across genomes than is possible by purely sequence-based methods, can be used for in silico screening of small molecule libraries for drug discovery, and can serve as aids in experimental structure determination by X-ray crystallography or cryo-EM. AF2 also turns out to be surprisingly good at predicting protein-protein and protein-peptide interactions. Nevertheless, while AF2 has clearly transformed protein structural bioinformatics, it is important to remember that not every model is high quality and limitations in relation to multi-domain proteins and protein dynamics remain.
Prof Dan Rigden is a biochemist turned bioinformatician who has 30 years’ experience in protein structure modelling. He has particular interest in using structural bioinformatics methods in X-ray crystallography and cryo-EM, as well as using structure predictions for the function annotation of cryptic proteins and protein families.