BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Halogenation Site-Selectivity Prediction Just Got Faster - Henriqu
 e Magri Marçon\, University of Cambridge
DTSTART:20250129T150000Z
DTEND:20250129T153000Z
UID:TALK226741@talks.cam.ac.uk
CONTACT:Lisa Masters
DESCRIPTION:Predicting aromatic substitution sites for new molecules remai
 n a challenge with large industry demand as its products have a myriad of 
 applications. Classical methods involve rule-based approaches to ab initio
  methods that scale in computational time for more complex scenarios of he
 teroaromatic and multi-substituted systems. Previous works have explored a
 b initio\, as well as hybrid methods with bespoke descriptors for each rea
 ction site (86% accuracy\, average 2\,899 ms/inference). Here\, we explore
  a data-driven model for halogenation site-selectivity achieving 80% accur
 acy with average 43 ms/inference. Our architecture combines machine learni
 ng with molecular fingerprints and algorithmic manipulation of chemical sc
 affolds. We also present an exploration of how different datasets – chlo
 rination\, bromination\, and iodination – can be combined into a superse
 t to increase prediction power of the final model. Finally\, model perform
 ance is higher when compared to chemist\, as they have through knowledge o
 f scaffolds they have previously worked with. This model compared to chemi
 sts. Although the sample size is small\, those working on the chemical ind
 ustry have deep knowledge on certain molecular scaffolds while fast and ac
 curate models can extend their reach to new areas.
LOCATION:Unilever Lecture Theatre\, Yusuf Hamied Department of Chemistry
END:VEVENT
END:VCALENDAR