BRADSHAW: A System for Automated Molecular Design – Application and Learnings
Stephen Pickett, Darren V.S. Green and Chris Luscombe
GlaxoSmithKline, Stevenage, UK
In the context of accelerating drug design to propose innovative new medicines, GSK has implemented a small molecule automated design platform called BRADSHAW [1]. This automated platform is able to facilitate the generation of new chemical ideas, the modelling and prediction of crucial properties, filtering, ranking and selection of novel compounds based on multiple objectives using active learning and other selection techniques. The final step is the annotation of results for discussion and dissemination. In this presentation we will describe the current status of BRADSHAW with several examples of its successful use in projects. We will also highlight and discuss potential solutions to several of the scientific challenges that have arisen through the application of BRADSHAW to real world problems. In particular, concepts of chemical space, how far is too far for an idea and how to identify the areas of space enriched in “good” compounds.
[1] Green, D.V.S., Pickett, S., Luscombe, C. et al. BRADSHAW: a system for automated molecular design. J Comput Aided Mol Des 34, 747–765 (2020)
