BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:OpenFold: Lessons learned and insights gained from rebuilding and 
 retraining AlphaFold2.    - Mohammed AlQuarishi\, Columbia University
DTSTART:20230522T130000Z
DTEND:20230522T140000Z
UID:TALK194656@talks.cam.ac.uk
CONTACT:Estelle Heather Pepper
DESCRIPTION:AlphaFold2 revolutionized structural biology by accurately pre
 dicting protein structures from sequence. Its implementation however (i) l
 acks the code and data required to train models for new tasks\, such as pr
 edicting alternate protein conformations or antibody structures\, (ii) is 
 unoptimized for commercially available computing hardware\, making large-s
 cale prediction campaigns impractical\, and (iii) remains poorly understoo
 d with respect to how training data and regimen influence accuracy. Here w
 e report OpenFold\, an optimized and trainable version of AlphaFold2. We t
 rain OpenFold from scratch and demonstrate that it fully reproduces AlphaF
 old2’s accuracy. By analyzing OpenFold training\, we find new relationsh
 ips between data size/diversity and prediction accuracy and gain insights 
 into how OpenFold learns to fold proteins during its training process.\n 
LOCATION:Zoom: https://zoom.us/j/94647233365?pwd=SGlubDdCL2ZSTWJvNHNjV0NiW
 HM5dz09
END:VEVENT
END:VCALENDAR