BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:State-of-the-art QSAR modelling with SOAP -  William McCorkindale
DTSTART:20200203T170000Z
DTEND:20200203T173000Z
UID:TALK139144@talks.cam.ac.uk
CONTACT:Bingqing Cheng
DESCRIPTION:Predicting the properties/bioactivity of a molecule from its s
 tructure is a longstanding challenge in drug discovery\, and in the past f
 ew years graph neural networks have found increasing popularity amongst th
 e cheminformatics community. Such models operate on atomic and/or bond fea
 tures over 2D graph descriptions of molecules\, and they have been establi
 shed as the current state-of-the-art.\n\nHowever\, many properties of inte
 rest (eg solvation\, binding affinity) involve 3D interactions between the
  local regions of molecules\, which might not be easily described by 2D de
 scriptors and motivates the use of a model which captures the 3D shape of 
 a molecule.\n\nIn this talk I will introduce a GP regression model which u
 ses the SOAP rematch kernel\, and show that it can beat state-of-the-art g
 raph neural networks on benchmark datasets. In addition\, I will describe 
 some recently proposed metrics for quantifying the quality of the uncertai
 nty predictions from ML models.\n
LOCATION:Mott Seminar (531) room\, top floor of the Mott Building\, in the
  Cavendish Laboratory\, West Cambridge.
END:VEVENT
END:VCALENDAR