06 May 2020
John Jumper (Google DeepMind)
Some of the ~27 proteins in the coronavirus genome are still uncharacterized in structure and/or function. Experimental protein characterization is a difficult and time-consuming process, so researchers have developed computational methods to predict protein properties solely from readily-available genetic sequences. This talk will describe our and others’ work to predict the 3-D structure of understudied proteins in the SARS-CoV-2 genome, as well as opportunities and challenges in using these structures to understand SARS-CoV-2. Our predicted protein structures are available here, created with an improved version of the AlphaFold algorithm and available under a Creative Commons open source license.