09-22, 11:25–11:55 (US/Eastern), Main Room
Aleksandra Kalisz
The talk will be a high-level overview of applications of machine learning to finding new drugs and curing diseases. The main focus will be on applications of reinforcement learning to generative molecule design methods. We’ll start by breaking down drug discovery into smaller problems that can be solved using machine learning. We’ll highlight the challenges often encountered when dealing with these problems. Finally, we’ll focus on generative molecule design and how reinforcement learning can be successfully applied to design drug candidates. The objective of this talk is to build a high-level understanding of how drug discovery can be automated using machine learning and reinforcement learning. The central thesis is that drug discovery can be viewed as a learning problem and AI can vastly help with finding new drugs faster. The key takeaways of the talk are listed below.
Key takeaways
1. Ways of applying ML to drug discovery and their challenges.
2. Understanding why we need Generative Molecule Design (GMD) methods.
3. GMD as a reinforcement learning problem.
We will specifically focus on ways we apply AI to drug discovery at Exscientia with the main emphasis on applications of reinforcement learning and how to use it successfully in this domain. The talk should be easy to follow for people with a STEM background. A basic understanding of reinforcement learning or drug discovery is a bonus but it’s not required.
Previous knowledge expected
Ola is a Senior Research Scientist and she currently leads the Reinforcement Learning team at Exscientia. Together with her team, she guides Exscientia's strategy for building new methods that can produce molecules which optimise a desired set of properties and can lead to finding new drug candidates. Prior to joining Exscientia Ola worked on different applications of Reinforcement Learning at the A*STAR institute in Singapore, California Institute of Technology as well as at the University of Edinburgh where she graduated.